def execute_running(self):
super().execute_running()
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
self.intercept = (constants.Field.OurGoalSegment.center() -
main.ball().pos) * 0.25 + main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of OpponentRobots sorted rudimentarily by their threat
# Right now, this is (inverse of) distance to the ball * 2 + y position.
# Needs tuning/improvement. Right now this is excessively defensive
sorted_opponents = sorted(
filter(lambda robot: robot != their_kicker, main.their_robots()),
key=lambda robot: self.calculate_shot_chance(robot),
reverse=True)
# Decide what each marking robot should do
# @sorted_opponents contains the robots we want to mark by priority
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if i < len(sorted_opponents):
mark_i.mark_robot = sorted_opponents[i]
self.marks[2].mark_robot = their_kicker
self.marks[2].mark_ratio = 0.5
def execute_running(self):
super().execute_running()
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of OpponentRobots sorted rudimentarily by their threat
# Right now, this is (inverse of) distance to the ball * 2 + y position.
# Needs tuning/improvement. Right now this is excessively defensive
sorted_opponents = sorted(
filter(lambda robot: robot != their_kicker, main.their_robots()),
key=lambda robot: robot.pos.dist_to(ball_pos) * 2 + robot.pos.y)
# Decide what each marking robot should do
# @sorted_opponents contains the robots we want to mark by priority
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if i < len(sorted_opponents):
mark_i.mark_robot = sorted_opponents[i]
def execute_running(self):
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(), key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of (OpponentRobot, float distToClosestOfOurBots) tuples
# we'll use these in a sec to assign our marking robots
open_opps_and_dists = []
for opp in main.their_robots():
# ignore their kicker
if opp == their_kicker: continue
ball_dist = opp.pos.dist_to(ball_pos)
# see if the opponent is close enough to the ball for us to care
# if it is, we record the closest distance from one of our robots to it
if ball_dist < 3.0:
# which of our robots is closest to this opponent
closest_self_dist = min([bot.pos.dist_to(opp.pos) for bot in main.our_robots()])
open_opps_and_dists.append( (opp, closest_self_dist) )
# Decide what each marking robot should do
# @open_opps contains the robots we want to mark (if any)
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if mark_i.robot != None:
if i < len(open_opps_and_dists):
# mark the opponent
mark_i.mark_robot = open_opps_and_dists[i][0]
else:
pass
# NOTE: the old code ran these motion commands INSTEAD of running the mark command
# we could do that, but it'd require removing the subbehavior and re-adding a move or something...
# # move towards the ball and face it, but don't get within field center's radius
# pos = mark_i.robot.pos
# target = pos + (ball_pos - pos).normalized() * (ball_pos.dist_to(pos) - constants.Field.CenterRadius)
# mark_i.robot.move(target)
# mark_i.face(ball_pos)
# tell the marking robots to avoid eachother more than normal
for i, mark_i in enumerate(self.marks):
for j, mark_j in enumerate(self.marks):
if i == j: continue
if mark_i.robot != None and mark_j.robot != None:
mark_i.robot.set_avoid_teammate_radius(mark_j.robot.shell_id(), 0.5)
def execute_running(self):
super().execute_running()
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of (OpponentRobot, float distToClosestOfOurBots) tuples
# we'll use these in a sec to assign our marking robots
open_opps_and_dists = []
for opp in main.their_robots():
# ignore their kicker
if opp == their_kicker: continue
ball_dist = opp.pos.dist_to(ball_pos)
# see if the opponent is close enough to the ball for us to care
# if it is, we record the closest distance from one of our robots to it
if ball_dist < 3.0:
# which of our robots is closest to this opponent
closest_self_dist = min(
[bot.pos.dist_to(opp.pos) for bot in main.our_robots()])
open_opps_and_dists.append((opp, closest_self_dist))
# Decide what each marking robot should do
# @open_opps contains the robots we want to mark (if any)
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if mark_i.robot != None:
if i < len(open_opps_and_dists):
# mark the opponent
mark_i.mark_robot = open_opps_and_dists[i][0]
else:
pass
# NOTE: the old code ran these motion commands INSTEAD of running the mark command
# we could do that, but it'd require removing the subbehavior and re-adding a move or something...
# # move towards the ball and face it, but don't get within field center's radius
# pos = mark_i.robot.pos
# target = pos + (ball_pos - pos).normalized() * (ball_pos.dist_to(pos) - constants.Field.CenterRadius)
# mark_i.robot.move(target)
# mark_i.face(ball_pos)
# tell the marking robots to avoid eachother more than normal
for i, mark_i in enumerate(self.marks):
for j, mark_j in enumerate(self.marks):
if i == j: continue
if mark_i.robot != None and mark_j.robot != None:
mark_i.robot.set_avoid_teammate_radius(
mark_j.robot.shell_id(), 0.5)
def can_collect_ball_before_opponent(our_robots_to_check=None,
their_robots_to_check=None,
our_robots_to_dodge=None,
their_robots_to_dodge=None,
valid_error_percent=0.05):
if our_robots_to_check is None:
our_robots_to_check = main.our_robots()
if their_robots_to_check is None:
their_robots_to_check = main.their_robots()
if our_robots_to_dodge is None:
our_robots_to_dodge = main.our_robots()
if their_robots_to_dodge is None:
their_robots_to_dodge = main.their_robots()
shortest_opp_time = float("inf")
shortest_our_time = float("inf")
dodge_dist = constants.Robot.Radius
closest_robot = None
# TODO: Do some sort of prediction as the ball moves
target_pos = main.ball().pos
# TODO: Take velocity and acceleration into account
# Find closest opponent robot
for bot in their_robots_to_check:
dist = estimate_path_length(bot.pos, target_pos, our_robots_to_dodge,
dodge_dist)
target_dir = (target_pos - bot.pos).normalized()
time = robocup.get_trapezoidal_time(
dist, dist, 2.2, 1,
target_dir.dot(bot.vel) / target_dir.mag(), 0)
if (time < shortest_opp_time):
shortest_opp_time = time
# Find closest robot on our team
for bot in our_robots_to_check:
dist = estimate_path_length(bot.pos, target_pos, their_robots_to_dodge,
dodge_dist)
target_dir = (target_pos - bot.pos).normalized()
time = robocup.get_trapezoidal_time(
dist, dist, 2.2, 1,
target_dir.dot(bot.vel) / target_dir.mag(), 0)
if (time < shortest_our_time):
shortest_our_time = time
closest_robot = bot
# Greater than 1 when we are further away
print(shortest_our_time)
print(shortest_opp_time)
return shortest_our_time < shortest_opp_time * (
1 + valid_error_percent), closest_robot
def can_collect_ball_before_opponent(our_robots_to_check=None,
their_robots_to_check=None,
our_robots_to_dodge=None,
their_robots_to_dodge=None,
valid_error_percent=0.05):
if our_robots_to_check is None:
our_robots_to_check = main.our_robots()
if their_robots_to_check is None:
their_robots_to_check = main.their_robots()
if our_robots_to_dodge is None:
our_robots_to_dodge = main.our_robots()
if their_robots_to_dodge is None:
their_robots_to_dodge = main.their_robots()
shortest_opp_time = float("inf")
shortest_our_time = float("inf")
dodge_dist = constants.Robot.Radius
closest_robot = None
# TODO: Do some sort of prediction as the ball moves
target_pos = main.ball().pos
# TODO: Take velocity and acceleration into account
# Find closest opponent robot
for bot in their_robots_to_check:
dist = estimate_path_length(bot.pos, target_pos, our_robots_to_dodge,
dodge_dist)
target_dir = (target_pos - bot.pos).normalized()
time = robocup.get_trapezoidal_time(dist, dist, 2.2, 1,
target_dir.dot(bot.vel) /
target_dir.mag(), 0)
if (time < shortest_opp_time):
shortest_opp_time = time
# Find closest robot on our team
for bot in our_robots_to_check:
dist = estimate_path_length(bot.pos, target_pos, their_robots_to_dodge,
dodge_dist)
target_dir = (target_pos - bot.pos).normalized()
time = robocup.get_trapezoidal_time(dist, dist, 2.2, 1,
target_dir.dot(bot.vel) /
target_dir.mag(), 0)
if (time < shortest_our_time):
shortest_our_time = time
closest_robot = bot
return shortest_our_time < shortest_opp_time * (1 + valid_error_percent
), closest_robot
def num_on_offense():
# Complementary filter based on...
# Distance to their goal
# Distance to the ball
goal_loc = robocup.Point(0, constants.Field.Length)
corner_loc = robocup.Point(constants.Field.Width / 2, 0)
ball_loc = main.ball().pos
max_goal_dis = (goal_loc - corner_loc).mag()
ball_to_goal = (goal_loc - ball_loc).mag()
offense_ctr = 0
filter_coeff = 0.7
score_cutoff = .3
# For each of their robots
for bot in main.their_robots():
if bot.visible:
dist_to_ball = (bot.pos - ball_loc).mag()
dist_to_goal = (bot.pos - goal_loc).mag()
goal_coeff = dist_to_goal / max_goal_dis
if ball_to_goal != 0:
ball_coeff = 1 - (dist_to_ball / ball_to_goal)
else:
ball_coeff = 1
ball_coeff = max(0, ball_coeff * ball_coeff)
score = filter_coeff * goal_coeff + (1 - filter_coeff) * ball_coeff
# Only count if their score is above the cutoff
if (score > score_cutoff):
offense_ctr += 1
return offense_ctr
def is_goalie_close(self):
# if any of there robots are in near chip range chip over them
close_check = False
for r in main.their_robots():
close_check = close_check or (
r.pos - main.ball().pos).mag() < self.goalie_range
return close_check
def in_chip_distance(self):
chipIndicator = False
for r in main.their_robots():
chipIndicator = chipIndicator or (
r.pos - constants.Field.OurGoalSegment.nearest_point(
r.pos)).mag() < self.chip_distance
return chipIndicator
def other_robot_touching_ball(self):
max_radius = constants.Robot.Radius + constants.Ball.Radius + 0.03
for bot in list(main.our_robots()) + list(main.their_robots()):
if bot.visible and (not bot.is_ours() or not bot.shell_id() == self.kicker_shell_id):
if bot.pos.near_point(main.ball().pos, max_radius):
return True
return False
def other_robot_touching_ball(self):
max_radius = constants.Robot.Radius + constants.Ball.Radius + 0.03
for bot in list(main.our_robots()) + list(main.their_robots()):
if bot.visible and (not bot.is_ours()
or not bot.shell_id() == self.kicker_shell_id):
if bot.pos.near_point(main.ball().pos, max_radius):
return True
return False
def other_robot_touching_ball(self):
for bot in filter(lambda bot: bot.visible,
list(main.our_robots()) + list(main.their_robots())):
if (bot.is_ours() and bot.has_ball()
and bot.shell_id() != self.kicker_shell_id):
return True
if ((bot.shell_id() != self.kicker_shell_id or not bot.is_ours())
and self._bot_in_radius(bot)):
return True
return False
def find_defense_positions(ignore_robots=[]):
their_risk_scores = []
for bot in main.their_robots():
score = estimate_risk_score(bot.pos, ignore_robots)
main.system_state().draw_text("Risk: " + str(int(score * 100)),
bot.pos, constants.Colors.White,
"Defense")
their_risk_scores.extend([score])
# Sorts bot array based on their score
zipped_array = zip(their_risk_scores, main.their_robots())
sorted_array = sorted(zipped_array, reverse=True)
sorted_bot = [bot for (scores, bot) in sorted_array]
area_def_pos = create_area_defense_zones(ignore_robots)
return area_def_pos, sorted_bot[0], sorted_bot[1]
def find_defense_positions(ignore_robots=[]):
their_risk_scores = []
for bot in main.their_robots():
score = estimate_risk_score(bot.pos, ignore_robots)
main.debug_drawer().draw_text("Risk: " + str(int(score * 100)),
bot.pos, constants.Colors.White,
"Defense")
their_risk_scores.extend([score])
# Sorts bot array based on their score
zipped_array = zip(their_risk_scores, main.their_robots())
sorted_array = sorted(zipped_array, reverse=True)
sorted_bot = [bot for (scores, bot) in sorted_array]
area_def_pos = create_area_defense_zones(ignore_robots)
return area_def_pos, sorted_bot[0], sorted_bot[1]
def other_robot_touching_ball(self):
for bot in filter(lambda bot: bot.visible,
list(main.our_robots()) + list(main.their_robots())):
if (bot.is_ours() and
bot.has_ball() and
bot.shell_id() != self.kicker_shell_id):
return True
if ((bot.shell_id() != self.kicker_shell_id or
not bot.is_ours()) and
self._bot_in_radius(bot)):
return True
return False
def execute_setup_penalty(self):
pt = robocup.Point(0, robocup.Field_PenaltyDist)
penalty_kicker = min(main.their_robots(), key=lambda r: (r.pos - pt).mag())
angle_rad = penalty_kicker.angle
shot_line = robocup.Line(penalty_kicker.pos, penalty_kicker.pos + robocup.Point.direction(angle_rad))
dest = shot_line.intersection(Goalie.RobotSegment)
if dest == None:
self.robot.move_to(robocup.Point(0, constants.Robot.Radius))
else:
dest.x = max(-Goalie.MaxX + constants.Robot.Radius, dest.x)
dest.y = min(Goalie.MaxX - constants.Robot.Radius, dest.x)
self.robot.move_to(dest)
def execute_running(self):
super().execute_running()
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of (OpponentRobot, float distToClosestOfOurBots) tuples
# we'll use these in a sec to assign our marking robots
open_opps_and_dists = []
for opp in main.their_robots():
# ignore their kicker
if opp == their_kicker: continue
ball_dist = opp.pos.dist_to(ball_pos)
# see if the opponent is close enough to the ball for us to care
# if it is, we record the closest distance from one of our robots to it
if ball_dist < 3.0:
# which of our robots is closest to this opponent
closest_self_dist = min(
[bot.pos.dist_to(opp.pos) for bot in main.our_robots()])
open_opps_and_dists.append((opp, closest_self_dist))
# Decide what each marking robot should do
# @open_opps contains the robots we want to mark (if any)
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if mark_i.robot != None:
if i < len(open_opps_and_dists):
# mark the opponent
mark_i.mark_robot = open_opps_and_dists[i][0]
else:
pass
def execute_running(self):
super().execute_running()
# abort if we can't see the ball
if not main.ball().valid:
return
ball_pos = main.ball().pos
# the closest of their bots to the ball is their kicker
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(ball_pos))
# we build an array of (OpponentRobot, float distToClosestOfOurBots) tuples
# we'll use these in a sec to assign our marking robots
open_opps_and_dists = []
for opp in main.their_robots():
# ignore their kicker
if opp == their_kicker: continue
ball_dist = opp.pos.dist_to(ball_pos)
# see if the opponent is close enough to the ball for us to care
# if it is, we record the closest distance from one of our robots to it
if ball_dist < 3.0:
# which of our robots is closest to this opponent
closest_self_dist = min([bot.pos.dist_to(opp.pos)
for bot in main.our_robots()])
open_opps_and_dists.append((opp, closest_self_dist))
# Decide what each marking robot should do
# @open_opps contains the robots we want to mark (if any)
# any robot that isn't assigned a mark_robot will move towards the ball
for i, mark_i in enumerate(self.marks):
if mark_i.robot != None:
if i < len(open_opps_and_dists):
# mark the opponent
mark_i.mark_robot = open_opps_and_dists[i][0]
else:
pass
def __init__(self):
super().__init__(continuous=True)
self.debug = False
self.add_transition(behavior.Behavior.State.start,
behavior.Behavior.State.running, lambda: True,
'immediately')
self.marks = []
if (main.game_state().is_their_direct()):
self.intercept = (constants.Field.OurGoalSegment.center() -
main.ball().pos) * 0.25 + main.ball().pos
self.add_subbehavior(skills.move.Move(self.intercept),
'intercept direct',
required=False,
priority=5)
for i in range(3):
mark_i = skills.mark.Mark()
mark_i.ratio = 0.7
self.add_subbehavior(mark_i,
'mark' + str(i),
required=False,
priority=3 - i)
self.marks.append(mark_i)
self.kick_eval = robocup.KickEvaluator(main.system_state())
self.kick_eval.debug = True
for i, robot in enumerate(main.our_robots()):
self.kick_eval.add_excluded_robot(robot)
for i, robot in enumerate(main.their_robots()):
self.kick_eval.add_excluded_robot(robot)
their_kicker = min(main.their_robots(),
key=lambda opp: opp.pos.dist_to(main.ball().pos))
self.kick_eval.add_excluded_robot(their_kicker)
def within_range(self):
shortest_opp_dist = 10
shortest_our_dist = 10
# TODO: Do some sort of prediction as the ball moves
target_pos = main.ball().pos
# Find closest opponent robot
for bot in main.their_robots():
dist = self.estimate_path_length(bot.pos, target_pos,
main.our_robots())
if (dist < shortest_opp_dist):
shortest_opp_dist = dist
# Find closest robot on our team
for bot in main.our_robots():
dist = self.estimate_path_length(bot.pos, target_pos,
main.their_robots())
if (dist < shortest_our_dist):
shortest_our_dist = dist
# Greater than 1 when we are further away
return shortest_our_dist < shortest_opp_dist * 1.05
def opponent_with_ball():
closest_bot, closest_dist = None, float("inf")
for bot in main.their_robots():
if bot.visible:
dist = (bot.pos - main.ball().pos).mag()
if dist < closest_dist:
closest_bot, closest_dist = bot, dist
if closest_bot == None:
return None
else:
if robot_has_ball(closest_bot):
return closest_bot
else:
return None
def opponent_with_ball():
closest_bot, closest_dist = None, float("inf")
for bot in main.their_robots():
if bot.visible:
dist = (bot.pos - main.ball().pos).mag()
if dist < closest_dist:
closest_bot, closest_dist = bot, dist
if closest_bot == None:
return None
else:
if robot_has_ball(closest_bot):
return closest_bot
else:
return None
def execute_setup_penalty(self):
pt = robocup.Point(0, robocup.Field_PenaltyDist)
penalty_kicker = min(main.their_robots(),
key=lambda r: (r.pos - pt).mag())
angle_rad = penalty_kicker.angle
shot_line = robocup.Line(
penalty_kicker.pos,
penalty_kicker.pos + robocup.Point.direction(angle_rad))
dest = shot_line.intersection(Goalie.RobotSegment)
if dest == None:
self.robot.move_to(robocup.Point(0, constants.Robot.Radius))
else:
dest.x = max(-Goalie.MaxX + constants.Robot.Radius, dest.x)
dest.y = min(Goalie.MaxX - constants.Robot.Radius, dest.x)
self.robot.move_to(dest)
def get_transition_point(self):
segment = constants.Field.OurGoalSegment
line = robocup.Line(main.ball().pos, main.their_robots()[0].pos)
goalie_pos = constants.Field.OurGoalSegment.center()
if (self.has_subbehavior_with_name('block')):
block_robot = self.subbehavior_with_name('block').robot
if (block_robot != None):
goalie_pos = block_robot.pos
#find the point that the robot is most likely going to shoot at
goal_intercept = segment.line_intersection(line)
#if robot is not lined up with the goal assume the bot will shoot center
if goal_intercept == None:
goal_intercept = segment.center()
#get the vector of the ball to the ideal shot point
ball_to_goal_intercept = goal_intercept - main.ball().pos
#normalize the vector
ball_to_goal_intercept = ball_to_goal_intercept.normalized()
#the ideal spot to be will be the chip distance times the normalized vector from the ball's
#current point.
ideal_defence = ball_to_goal_intercept * self.chip_distance + main.ball(
).pos
#find the line of the shot the robot will make
ball_to_goal_segment = robocup.Segment(main.ball().pos, goal_intercept)
#this is the point to get our robot to block the shot the quickest.
fastest_defence = ball_to_goal_segment.nearest_point(goalie_pos)
#if robot is blocking the shot already just go to the ideal point otherwise average the vectors
#based on the blocking percentage.
if (goalie_pos.dist_to(fastest_defence) <
(constants.Robot.Radius / 4)):
move_to_point = ideal_defence
else:
move_to_point = robocup.Point(
(ideal_defence.x * (self.block_percentage)) +
(fastest_defence.x * (1 - self.block_percentage)),
((ideal_defence.y * self.block_percentage) +
fastest_defence.y * (1 - self.block_percentage)))
return move_to_point
def get_closest_opponent(pos, direction_weight=0, excluded_robots=[]):
closest_bot, closest_dist = None, float("inf")
for bot in main.their_robots():
if bot.visible and bot not in excluded_robots:
dist = (bot.pos - pos).mag()
if (pos.y <= bot.pos.y):
dist *= (2 - direction_weight / 2)
else:
dist *= (2 + direction_weight / 2)
if dist < closest_dist:
closest_bot = bot
closest_dist = dist
return closest_bot
def space_coeff_at_pos(pos, excluded_robots=[], robots=None):
# TODO: Add in velocity prediction
if robots == None:
robots = main.their_robots()
max_dist = robocup.Point(constants.Field.Width / 2,
constants.Field.Length).mag()
total = 0
sensitivity = 8
for bot in robots:
if bot.visible and bot not in excluded_robots:
u = sensitivity * (bot.pos - pos).mag() / max_dist
# Use Triweight kernal function (Force to be positive)
# Graph looks much like a normal distribution
total += max((35 / 32) * pow((1 - pow(u, 2)), 3), 0)
return min(total, 1)
def execute_setup_penalty(self):
pt = robocup.Point(0, constants.Field.PenaltyDist)
penalty_kicker = min(main.their_robots(),
key=lambda r: (r.pos - pt).mag())
angle_rad = penalty_kicker.angle
shot_line = robocup.Line(
penalty_kicker.pos,
penalty_kicker.pos + robocup.Point.direction(angle_rad))
dest = Goalie.RobotSegment.line_intersection(shot_line)
if dest is None:
self.robot.move_to(
robocup.Point(0, constants.Robot.Radius + Goalie.OFFSET))
else:
dest.x = max(-Goalie.MaxX + constants.Robot.Radius, dest.x)
dest.x = min(Goalie.MaxX - constants.Robot.Radius, dest.x)
# Shots don't follow the top threat, they follow the inverse
# FIXME this is kind of a hack
dest.x = -dest.x
self.robot.move_to(dest)
def classify_opponent_robots(self):
del self.wingers[:]
del self.forwards[:]
for bot in main.their_robots():
if bot.visible and bot.pos.y < constants.Field.Length / 2:
robot_risk_score = self.calculate_robot_risk_scores(bot)
area_risk_score = self.calculate_area_risk_scores(bot)
features = [robot_risk_score, area_risk_score]
is_wing, class_score = evaluation.linear_classification.binary_classification(
features, WingerWall.WING_FORWARD_WEIGHTS,
WingerWall.WING_FORWARD_BIAS,
WingerWall.WING_FORWARD_CUTOFF)
is_wing = not is_wing #appears tobe inverted fix TODO
if is_wing:
self.wingers.append((class_score, bot))
else:
self.forwards.append((class_score, bot))
def opp_robot_blocking(self):
if (self.robot is None):
return False
# Closest opp robot in any direction
# To us, not the ball
closest_opp_robot = None
closest_opp_dist = float("inf")
for r in main.their_robots():
if ((r.pos - self.robot.pos).mag() < closest_opp_dist):
closest_opp_robot = r
closest_opp_dist = (r.pos - self.robot.pos).mag()
# Only do this if a robot is in range
robot_in_range = closest_opp_dist < .2 + 2 * constants.Robot.Radius
aim_dir = robocup.Point.direction(self.robot.angle)
robot_dir = (closest_opp_robot.pos - self.robot.pos)
# Only trigger if they are infront of us
robot_in_front = aim_dir.dot(robot_dir) > 0
closest_pt = robocup.Line(
self.robot.pos,
self.robot.pos + aim_dir).nearest_point(closest_opp_robot.pos)
does_hit_robot = (closest_opp_robot.pos - closest_pt
).mag() < constants.Robot.Radius
facing_their_side = robocup.Point.direction(self.robot.angle).y > 0
ret = (facing_their_side and robot_in_range and robot_in_front and
does_hit_robot)
if ret:
print("Panic kick")
main.debug_drawer().draw_text('panic kick', self.robot.pos,
(255, 255, 255), 'PivotKick')
return ret
def execute_setup_penalty(self):
if main.ball().valid:
pt = main.ball().pos
else:
# FIXME is this correct?
pt = robocup.Point(0, constants.Field.PenaltyLongDist)
penalty_kicker = min(main.their_robots(),
key=lambda r: (r.pos - pt).mag())
angle_rad = penalty_kicker.angle
shot_line = robocup.Line(
penalty_kicker.pos,
penalty_kicker.pos + robocup.Point.direction(angle_rad))
dest = Goalie.RobotSegment.line_intersection(shot_line)
if dest is None:
self.robot.move_to(
robocup.Point(0, constants.Robot.Radius + Goalie.OFFSET))
else:
dest.x = max(-Goalie.MaxX + constants.Robot.Radius, dest.x)
dest.x = min(Goalie.MaxX - constants.Robot.Radius, dest.x)
# Shots don't follow the top threat, they follow the inverse
# FIXME this is kind of a hack
dest.x = -dest.x
self.robot.move_to(dest)
def eval_pt_to_seg(self, origin, target):
end = target.delta().magsq()
# if target is a zero-length segment, there are no windows
if end == 0:
return [], None
if self.debug:
main.system_state().draw_line(target, constants.Colors.Blue, "Debug")
windows = [Window(0, end)]
# apply the obstacles
bots = filter(lambda bot: bot not in self.excluded_robots and bot.visible, (list(main.our_robots()) + list(main.their_robots())))
bot_locations = list(map(lambda bot: bot.pos, bots))
bot_locations.extend(self.hypothetical_robot_locations)
for pos in bot_locations:
d = (pos - origin).mag()
# whether or not we can chip over this bot
chip_overable = (self.chip_enabled
and (d < self.max_chip_range - constants.Robot.Radius)
and (d > self.min_chip_range + constants.Robot.Radius))
if not chip_overable:
self.obstacle_robot(windows, origin, target, pos)
# set the segment and angles for each window
p0 = target.get_pt(0)
delta = target.delta() / end
for w in windows:
w.segment = robocup.Segment(p0 + delta * w.t0, p0 + delta * w.t1)
w.a0 = (w.segment.get_pt(0) - origin).angle() * constants.RadiansToDegrees
w.a1 = (w.segment.get_pt(1) - origin).angle() * constants.RadiansToDegrees
best = max(windows, key=lambda w: w.segment.delta().magsq()) if len(windows) > 0 else None
if self.debug and best is not None:
main.system_state().draw_line(best.segment, constants.Colors.Green, "Debug")
main.system_state().draw_line(robocup.Line(origin, best.segment.center()), constants.Colors.Green, "Debug")
# # draw the windows if we're in debug mode
# if self.debug:
# for w in windows:
# pts = [origin, w.segment.get_pt(0), w.segment.get_pt(1)]
# color = (255, 0, 0) if w == best else (0, 255, 0)
# main.system_state().draw_polygon(pts, 3, color, "Windows")
return windows, best
def recalculate(self):
goalie = self.subbehavior_with_name('goalie')
defender1 = self.subbehavior_with_name('defender1')
defender2 = self.subbehavior_with_name('defender2')
behaviors = [goalie, defender1, defender2]
# if we don't have any bots to work with, don't waste time calculating
if all(bhvr.robot == None for bhvr in behaviors):
return
# A threat to our goal - something we'll actively defend against
class Threat:
def __init__(self, source=None):
self.source = source
self.ball_acquire_chance = 1.0
self.shot_chance = 1.0
self.assigned_handlers = []
self.best_shot_window = None
# an OpponentRobot or Point
@property
def source(self):
return self._source
@source.setter
def source(self, value):
self._source = value
# our source can be a Point or an OpponentRobot, this method returns the location of it
@property
def pos(self):
if self.source != None:
return self.source if isinstance(self.source, robocup.Point) else self.source.pos
# a list of our behaviors that will be defending against this threat
# as of now only Defender and Goalie
@property
def assigned_handlers(self):
return self._assigned_handlers
@assigned_handlers.setter
def assigned_handlers(self, value):
self._assigned_handlers = value
# our estimate of the chance that this threat will acquire the ball
# 1.0 if it already has it
# otherwise, a value from 0 to 1 gauging its likelihood to receive a pass
@property
def ball_acquire_chance(self):
return self._ball_acquire_chance
@ball_acquire_chance.setter
def ball_acquire_chance(self, value):
self._ball_acquire_chance = value
# our estimate of the chance of this threat making its shot on the goal given that it gets/has the ball
# NOTE: this is calculated excluding all of our robots on the field as obstacles
@property
def shot_chance(self):
return self._shot_chance
@shot_chance.setter
def shot_chance(self, value):
self._shot_chance = value
# his best window on our goal
@property
def best_shot_window(self):
return self._best_shot_window
@best_shot_window.setter
def best_shot_window(self, value):
self._best_shot_window = value
# our assessment of the risk of this threat
# should be between 0 and 1
@property
def score(self):
return self.ball_acquire_chance * self.shot_chance
# available behaviors we have to assign to threats
# only look at ones that have robots
# as we handle threats, we remove the handlers from this list
unused_threat_handlers = list(filter(lambda bhvr: bhvr.robot != None, [goalie, defender1, defender2]))
def set_block_lines_for_threat_handlers(threat):
if len(threat.assigned_handlers) == 0:
return
# make sure goalie is in the middle
if len(threat.assigned_handlers) > 1:
if goalie in threat.assigned_handlers:
idx = threat.assigned_handlers.index(goalie)
if idx != 1:
del threat.assigned_handlers[idx]
threat.assigned_handlers.insert(1, goalie)
if threat.best_shot_window != None:
center_line = robocup.Line(threat.pos, threat.best_shot_window.segment.center())
else:
center_line = robocup.Line(threat.pos, constants.Field.OurGoalSegment.center())
# find the angular width that each defender can block. We then space these out accordingly
angle_widths = []
for handler in threat.assigned_handlers:
dist_from_threat = handler.robot.pos.dist_to(threat.pos)
w = min(2.0 * math.atan2(constants.Robot.Radius, dist_from_threat), 0.15)
angle_widths.append(w)
# start on one edge of our available angle coverage and work counter-clockwise,
# assigning block lines to the bots as we go
spacing = 0.01 if len(threat.assigned_handlers) < 3 else 0.0 # spacing between each bot in radians
total_angle_coverage = sum(angle_widths) + (len(angle_widths) - 1)*spacing
start_vec = center_line.delta().normalized()
start_vec.rotate(robocup.Point(0,0), -total_angle_coverage / 2.0)
for i in range(len(angle_widths)):
handler = threat.assigned_handlers[i]
w = angle_widths[i]
start_vec.rotate(robocup.Point(0,0), w/2.0)
handler.block_line = robocup.Line(threat.pos, threat.pos + start_vec * 10)
start_vec.rotate(robocup.Point(0,0), w/2.0 + spacing)
def recalculate_threat_shot(threat_index):
if not isinstance(threat_index, int):
raise TypeError("threat_index should be an int")
# ignore all of our robots
excluded_robots = list(main.our_robots())
# behaviors before this threat are counted as obstacles in their TARGET position (where we've
# assigned them to go, not where they are right now)
hypothetical_obstacles = []
for t in threats[0:threat_index]:
hypothetical_obstacles.extend(map(lambda bhvr: bhvr.move_target, t.assigned_handlers))
threat = threats[threat_index]
threat.shot_chance, threat.best_shot_window = evaluation.shot.eval_shot(
pos=threat.pos,
target=constants.Field.OurGoalSegment,
windowing_excludes=excluded_robots,
hypothetical_robot_locations=[],
debug=False)
threats = []
# secondary threats are those that are somewhat close to our goal and open for a pass
# if they're farther than this down the field, we don't consider them threats
threat_max_y = constants.Field.Length / 2.0
potential_threats = [opp for opp in main.their_robots() if opp.pos.y < threat_max_y]
# find the primary threat
# if the ball is not moving OR it's moving towards our goal, it's the primary threat
# if it's moving, but not towards our goal, the primary threat is the robot on their team most likely to catch it
if main.ball().vel.mag() > 0.4:
# the line the ball's moving along
ball_travel_line = robocup.Line(main.ball().pos, main.ball().pos + main.ball().vel)
# this is a shot on the goal!
if evaluation.ball.is_moving_towards_our_goal():
ball_threat = Threat(main.ball().pos)
ball_threat.ball_acquire_chance = 1.0
ball_threat.shot_chance = 1.0
threats.append(ball_threat)
else:
# Check for a bot that's about to capture this ball and potentially shoot on the goal
# potential_receivers is an array of (OpponentRobot, angle) tuples, where the angle
# is the angle between the ball_travel_line and the line from the ball to the opponent
# bot - this is our metric for receiver likeliness.
potential_receivers = []
for opp in potential_threats:
# see if the bot is in the direction the ball is moving
if (opp.pos - ball_travel_line.get_pt(0)).dot(ball_travel_line.delta()) > 0:
# calculate the angle and add it to the list if it's within reason
nearest_pt = ball_travel_line.nearest_point(opp.pos)
dx = (nearest_pt - main.ball().pos).mag()
dy = (opp.pos - nearest_pt).mag()
angle = abs(math.atan2(dy, dx))
if angle < math.pi / 4.0:
potential_receivers.append( (opp, 1.0) )
# choose the receiver with the smallest angle from the ball travel line
if len(potential_receivers) > 0:
best_receiver_tuple = min(potential_receivers, key=lambda rcrv_tuple: rcrv_tuple[1])
if best_receiver_tuple != None:
receiver_threat = Threat(best_receiver_tuple[0])
receiver_threat.ball_acquire_chance = 0.9 # note: this value is arbitrary
receiver_threat.shot_chance = 0.9 # FIXME: calculate this
threats.append(receiver_threat)
else:
ball_threat = Threat(main.ball().pos)
ball_threat.ball_acquire_chance = 1.0
ball_threat.shot_chance = 0.9
threats.append(ball_threat)
else:
# primary threat is the ball or the opponent holding it
opp_with_ball = evaluation.ball.opponent_with_ball()
threat = Threat(opp_with_ball if opp_with_ball != None else main.ball().pos)
threat.ball_acquire_chance = 1.0
threat.shot_chance = 1.0 # FIXME: calculate, don't use 1.0
threats.append(threat)
# if an opponent has the ball or is potentially about to receive the ball,
# we look at potential receivers of it as threats
if isinstance(threats[0].source, robocup.OpponentRobot):
for opp in filter(lambda t: t.visible, potential_threats):
pass_chance = evaluation.passing.eval_pass(main.ball().pos, opp.pos, excluded_robots=[opp])
# give it a small chance because the obstacles in the way could move soon and we don't want to consider it a zero threatos, )
if pass_chance < 0.001: pass_chance = 0.4
# record the threat
threat = Threat(opp)
threat.ball_acquire_chance = pass_chance
threats.append(threat)
# Now we evaluate this opponent's shot on the goal
# exclude robots that have already been assigned to handle other threats
threat.shot_chance, threat.best_shot_window = evaluation.shot.eval_shot(
pos=opp.pos,
target=constants.Field.OurGoalSegment,
windowing_excludes=map(lambda bhvr: bhvr.robot, unused_threat_handlers),
debug=False)
if threat.shot_chance == 0:
# gve it a small chance because the shot could clear up a bit later and we don't want to consider it a zero threat
threat.shot_chance = 0.2
else:
# the ball isn't possessed by an opponent, so we just look at opponents with shots on the goal
for opp in potential_threats:
# record the threat
lurker = Threat(opp)
lurker.ball_acquire_chance = 0.5 # note: this is a bullshit value
threats.append(lurker)
recalculate_threat_shot(len(threats) - 1)
# only consider the top three threats
threats.sort(key=lambda threat: threat.score, reverse=True)
threats = threats[0:3]
# print("sorted threats:")
# for idx, t in enumerate(threats):
# print("t[" + str(idx) + "]: " + str(t.source) + "shot: " + str(t.shot_chance) + "; pass:"******"; score:" + str(t.score))
# print("sorted threat scores: " + str(list(map(lambda t: str(t.score) + " " + str(t.source), threats))))
# print("Unused handlers: " + str(unused_threat_handlers))
# print("---------------------")
smart = False
if not smart:
# only deal with top two threats
threats_to_block = threats[0:2]
# print('threats to block: ' + str(list(map(lambda t: t.source, threats_to_block))))
threat_idx = 0
while len(unused_threat_handlers) > 0:
threats_to_block[threat_idx].assigned_handlers.append(unused_threat_handlers[0])
del unused_threat_handlers[0]
threat_idx = (threat_idx + 1) % len(threats_to_block)
for t_idx, t in enumerate(threats_to_block):
recalculate_threat_shot(t_idx)
set_block_lines_for_threat_handlers(t)
else:
# assign all of our defenders to do something
while len(unused_threat_handlers) > 0:
# prioritize by threat score, highest first
top_threat = max(threats, key=lambda threat: threat.score)
# assign the next handler to this threat
handler = unused_threat_handlers[0]
top_threat.assigned_handlers.append(handler)
del unused_threat_handlers[0]
# reassign the block line for each handler of this threat
set_block_lines_for_threat_handlers(top_threat)
# recalculate the shot now that we have
recalculate_threat_shot(0)
# tell the bots where to move / what to block and draw some debug stuff
for idx, threat in enumerate(threats):
# recalculate, including all current bots
# FIXME: do we want this?
# recalculate_threat_shot(idx)
# the line they'll be shooting down/on
if threat.best_shot_window != None:
shot_line = robocup.Segment(threat.pos, threat.best_shot_window.segment.center())
else:
shot_line = robocup.Segment(threat.pos, robocup.Point(0, 0))
# debug output
if self.debug:
for handler in threat.assigned_handlers:
# handler.robot.add_text("Marking: " + str(threat.source), constants.Colors.White, "Defense")
main.system_state().draw_circle(handler.move_target, 0.02, constants.Colors.Blue, "Defense")
# draw some debug stuff
if threat.best_shot_window != None:
# draw shot triangle
pts = [threat.pos, threat.best_shot_window.segment.get_pt(0), threat.best_shot_window.segment.get_pt(1)]
shot_color = (255, 0, 0, 150) # translucent red
main.system_state().draw_polygon(pts, shot_color, "Defense")
main.system_state().draw_line(threat.best_shot_window.segment, constants.Colors.Red, "Defense")
chance, best_window = evaluation.shot.eval_shot(threat.pos, constants.Field.OurGoalSegment)
main.system_state().draw_text("Shot: " + str(int(threat.shot_chance * 100.0)) + "% / " + str(int(chance*100)) + "%", shot_line.center(), constants.Colors.White, "Defense")
# draw pass lines
if idx > 0:
pass_line = robocup.Segment(main.ball().pos, threat.pos)
main.system_state().draw_line(pass_line, constants.Colors.Red, "Defense")
main.system_state().draw_text("Pass: "******"%", pass_line.center(), constants.Colors.White, "Defense")
def get_threat_list(self, unused_threat_handlers):
# List of (position, score, Robot/None)
threats = []
potential_threats = main.their_robots()
# find the primary threat
# if the ball is not moving OR it's moving towards our goal, it's the primary threat
# if it's moving, but not towards our goal, the primary threat is the robot on their team most likely to catch it
if (main.ball().vel.mag() > 0.4):
if evaluation.ball.is_moving_towards_our_goal():
# Add tuple of pos and score
threats.append((main.ball().pos, 1, None))
else:
# Get all potential receivers
potential_receivers = []
for opp in potential_threats:
if self.estimate_potential_recievers_score(opp) == 1:
potential_receivers.append((opp.pos, 1, opp))
if len(potential_receivers) > 0:
# Add best receiver to threats
# TODO Calc shot chance
best_tuple = min(potential_receivers,
key=lambda rcrv_tuple: rcrv_tuple[1])
threats.append((best_tuple[0], .81, best_tuple[2]))
else:
# Just deal with ball if no recievers
threats.append((main.ball().pos, .9, None))
else:
# Assume opp is dribbling ball
if not constants.Field.OurGoalZoneShape.contains_point(
main.ball().pos):
# TODO: Calc shot chance
threats.append((main.ball().pos, 1, None))
# if there are threats, check pass and shot chances
# If the first item is not a ball, it is most likely a pass
if len(threats) > 0 and threats[0][0] != main.ball().pos:
for opp in potential_threats:
# Exclude robots that have been assigned already
excluded_bots = []
for r in map(lambda bhvr: bhvr.robot, unused_threat_handlers):
excluded_bots.append(r)
threats.append((opp.pos, self.estimate_risk_score(
opp, excluded_bots), opp))
else:
for opp in potential_threats:
# Exclude all robots
self.kick_eval.excluded_robots.clear()
self.kick_eval.add_excluded_robot(opp)
for r in main.our_robots():
self.kick_eval.add_excluded_robot(r)
point, shotChance = self.kick_eval.eval_pt_to_our_goal(opp.pos)
# Note: 0.5 is a bullshit value
threats.append((opp.pos, 0.5 * shotChance, opp))
# Prevent threats from being below our goal line (causes incorrect pos)
def _adjust_pt(threat):
pt = threat[0]
pt.y = max(pt.y, 0.1)
return (pt,) + threat[1:]
threats = list(map(_adjust_pt, threats))
return threats
def on_enter_running(self): b = skills.mark.Mark() b.mark_robot = main.their_robots()[0] self.add_subbehavior(b, name='mark', required=True)
def recalculate(self):
goalie = self.subbehavior_with_name('goalie')
defender1 = self.subbehavior_with_name('defender1')
defender2 = self.subbehavior_with_name('defender2')
behaviors = [goalie, defender1, defender2]
# if we don't have any bots to work with, don't waste time calculating
if all(bhvr.robot == None for bhvr in behaviors):
return
# A threat to our goal - something we'll actively defend against
class Threat:
def __init__(self, source=None):
self.source = source
self.ball_acquire_chance = 1.0
self.shot_chance = 1.0
self.assigned_handlers = []
self.best_shot_window = None
# an OpponentRobot or Point
@property
def source(self):
return self._source
@source.setter
def source(self, value):
self._source = value
# our source can be a Point or an OpponentRobot, this method returns the location of it
@property
def pos(self):
if self.source != None:
return self.source if isinstance(
self.source, robocup.Point) else self.source.pos
# a list of our behaviors that will be defending against this threat
# as of now only Defender and Goalie
@property
def assigned_handlers(self):
return self._assigned_handlers
@assigned_handlers.setter
def assigned_handlers(self, value):
self._assigned_handlers = value
# our estimate of the chance that this threat will acquire the ball
# 1.0 if it already has it
# otherwise, a value from 0 to 1 gauging its likelihood to receive a pass
@property
def ball_acquire_chance(self):
return self._ball_acquire_chance
@ball_acquire_chance.setter
def ball_acquire_chance(self, value):
self._ball_acquire_chance = value
# our estimate of the chance of this threat making its shot on the goal given that it gets/has the ball
# NOTE: this is calculated excluding all of our robots on the field as obstacles
@property
def shot_chance(self):
return self._shot_chance
@shot_chance.setter
def shot_chance(self, value):
self._shot_chance = value
# his best window on our goal
@property
def best_shot_window(self):
return self._best_shot_window
@best_shot_window.setter
def best_shot_window(self, value):
self._best_shot_window = value
# our assessment of the risk of this threat
# should be between 0 and 1
@property
def score(self):
return self.ball_acquire_chance * self.shot_chance
# available behaviors we have to assign to threats
# only look at ones that have robots
# as we handle threats, we remove the handlers from this list
unused_threat_handlers = list(
filter(lambda bhvr: bhvr.robot != None,
[goalie, defender1, defender2]))
def set_block_lines_for_threat_handlers(threat):
if len(threat.assigned_handlers) == 0:
return
# make sure goalie is in the middle
if len(threat.assigned_handlers) > 1:
if goalie in threat.assigned_handlers:
idx = threat.assigned_handlers.index(goalie)
if idx != 1:
del threat.assigned_handlers[idx]
threat.assigned_handlers.insert(1, goalie)
if threat.best_shot_window != None:
center_line = robocup.Line(
threat.pos, threat.best_shot_window.segment.center())
else:
center_line = robocup.Line(
threat.pos, constants.Field.OurGoalSegment.center())
# find the angular width that each defender can block. We then space these out accordingly
angle_widths = []
for handler in threat.assigned_handlers:
dist_from_threat = handler.robot.pos.dist_to(threat.pos)
w = min(
2.0 * math.atan2(constants.Robot.Radius, dist_from_threat),
0.15)
angle_widths.append(w)
# start on one edge of our available angle coverage and work counter-clockwise,
# assigning block lines to the bots as we go
spacing = 0.01 if len(
threat.assigned_handlers
) < 3 else 0.0 # spacing between each bot in radians
total_angle_coverage = sum(angle_widths) + (len(angle_widths) -
1) * spacing
start_vec = center_line.delta().normalized()
start_vec.rotate(robocup.Point(0, 0), -total_angle_coverage / 2.0)
for i in range(len(angle_widths)):
handler = threat.assigned_handlers[i]
w = angle_widths[i]
start_vec.rotate(robocup.Point(0, 0), w / 2.0)
handler.block_line = robocup.Line(threat.pos,
threat.pos + start_vec * 10)
start_vec.rotate(robocup.Point(0, 0), w / 2.0 + spacing)
def recalculate_threat_shot(threat_index):
if not isinstance(threat_index, int):
raise TypeError("threat_index should be an int")
# ignore all of our robots
excluded_robots = list(main.our_robots())
# behaviors before this threat are counted as obstacles in their TARGET position (where we've
# assigned them to go, not where they are right now)
hypothetical_obstacles = []
for t in threats[0:threat_index]:
hypothetical_obstacles.extend(
map(lambda bhvr: bhvr.move_target, t.assigned_handlers))
threat = threats[threat_index]
self.win_eval.excluded_robots.clear()
for r in excluded_robots:
self.win_eval.add_excluded_robot(r)
_, threat.best_shot_window = self.win_eval.eval_pt_to_our_goal(
threat.pos)
if threat.best_shot_window is not None:
threat.shot_chance = threat.best_shot_window.shot_success
else:
threat.shot_chance = 0.0
threats = []
# secondary threats are those that are somewhat close to our goal and open for a pass
# if they're farther than this down the field, we don't consider them threats
threat_max_y = constants.Field.Length / 2.0
potential_threats = [
opp for opp in main.their_robots() if opp.pos.y < threat_max_y
]
# find the primary threat
# if the ball is not moving OR it's moving towards our goal, it's the primary threat
# if it's moving, but not towards our goal, the primary threat is the robot on their team most likely to catch it
if main.ball().vel.mag() > 0.4:
# the line the ball's moving along
ball_travel_line = robocup.Line(main.ball().pos,
main.ball().pos + main.ball().vel)
# this is a shot on the goal!
if evaluation.ball.is_moving_towards_our_goal():
ball_threat = Threat(main.ball().pos)
ball_threat.ball_acquire_chance = 1.0
ball_threat.shot_chance = 1.0
threats.append(ball_threat)
else:
# Check for a bot that's about to capture this ball and potentially shoot on the goal
# potential_receivers is an array of (OpponentRobot, angle) tuples, where the angle
# is the angle between the ball_travel_line and the line from the ball to the opponent
# bot - this is our metric for receiver likeliness.
potential_receivers = []
for opp in potential_threats:
# see if the bot is in the direction the ball is moving
if (opp.pos - ball_travel_line.get_pt(0)).dot(
ball_travel_line.delta()) > 0:
# calculate the angle and add it to the list if it's within reason
nearest_pt = ball_travel_line.nearest_point(opp.pos)
dx = (nearest_pt - main.ball().pos).mag()
dy = (opp.pos - nearest_pt).mag()
angle = abs(math.atan2(dy, dx))
if angle < math.pi / 4.0:
potential_receivers.append((opp, 1.0))
# choose the receiver with the smallest angle from the ball travel line
if len(potential_receivers) > 0:
best_receiver_tuple = min(
potential_receivers,
key=lambda rcrv_tuple: rcrv_tuple[1])
if best_receiver_tuple != None:
receiver_threat = Threat(best_receiver_tuple[0])
receiver_threat.ball_acquire_chance = 0.9 # note: this value is arbitrary
receiver_threat.shot_chance = 0.9 # FIXME: calculate this
threats.append(receiver_threat)
else:
ball_threat = Threat(main.ball().pos)
ball_threat.ball_acquire_chance = 1.0
ball_threat.shot_chance = 0.9
threats.append(ball_threat)
else:
# primary threat is the ball or the opponent holding it
opp_with_ball = evaluation.ball.opponent_with_ball()
threat = Threat(
opp_with_ball if opp_with_ball != None else main.ball().pos)
threat.ball_acquire_chance = 1.0
threat.shot_chance = 1.0 # FIXME: calculate, don't use 1.0
threats.append(threat)
# if an opponent has the ball or is potentially about to receive the ball,
# we look at potential receivers of it as threats
if isinstance(threats[0].source, robocup.OpponentRobot):
for opp in filter(lambda t: t.visible, potential_threats):
pass_chance = evaluation.passing.eval_pass(
main.ball().pos, opp.pos, excluded_robots=[opp])
# give it a small chance because the obstacles in the way could move soon and we don't want to consider it a zero threatos, )
if pass_chance < 0.001: pass_chance = 0.4
# record the threat
threat = Threat(opp)
threat.ball_acquire_chance = pass_chance
threats.append(threat)
# Now we evaluate this opponent's shot on the goal
# exclude robots that have already been assigned to handle other threats
self.win_eval.excluded_robots.clear()
for r in map(lambda bhvr: bhvr.robot, unused_threat_handlers):
self.win_eval.add_excluded_robot(r)
_, threat.best_shot_window = self.win_eval.eval_pt_to_our_goal(
opp.pos)
if threat.best_shot_window is not None:
threat.shot_chance = threat.best_shot_window.shot_success
else:
threat.shot_chance = 0.0
if threat.shot_chance == 0:
# gve it a small chance because the shot could clear up a bit later and we don't want to consider it a zero threat
threat.shot_chance = 0.2
else:
# the ball isn't possessed by an opponent, so we just look at opponents with shots on the goal
for opp in potential_threats:
# record the threat
lurker = Threat(opp)
lurker.ball_acquire_chance = 0.5 # note: this is a bullshit value
threats.append(lurker)
recalculate_threat_shot(len(threats) - 1)
# only consider the top three threats
threats.sort(key=lambda threat: threat.score, reverse=True)
threats = threats[0:3]
# print("sorted threats:")
# for idx, t in enumerate(threats):
# print("t[" + str(idx) + "]: " + str(t.source) + "shot: " + str(t.shot_chance) + "; pass:"******"; score:" + str(t.score))
# print("sorted threat scores: " + str(list(map(lambda t: str(t.score) + " " + str(t.source), threats))))
# print("Unused handlers: " + str(unused_threat_handlers))
# print("---------------------")
smart = False
if not smart:
# only deal with top two threats
threats_to_block = threats[0:2]
# print('threats to block: ' + str(list(map(lambda t: t.source, threats_to_block))))
threat_idx = 0
while len(unused_threat_handlers) > 0:
threats_to_block[threat_idx].assigned_handlers.append(
unused_threat_handlers[0])
del unused_threat_handlers[0]
threat_idx = (threat_idx + 1) % len(threats_to_block)
for t_idx, t in enumerate(threats_to_block):
recalculate_threat_shot(t_idx)
set_block_lines_for_threat_handlers(t)
else:
# assign all of our defenders to do something
while len(unused_threat_handlers) > 0:
# prioritize by threat score, highest first
top_threat = max(threats, key=lambda threat: threat.score)
# assign the next handler to this threat
handler = unused_threat_handlers[0]
top_threat.assigned_handlers.append(handler)
del unused_threat_handlers[0]
# reassign the block line for each handler of this threat
set_block_lines_for_threat_handlers(top_threat)
# recalculate the shot now that we have
recalculate_threat_shot(0)
# tell the bots where to move / what to block and draw some debug stuff
for idx, threat in enumerate(threats):
# recalculate, including all current bots
# FIXME: do we want this?
# recalculate_threat_shot(idx)
# the line they'll be shooting down/on
if threat.best_shot_window != None:
shot_line = robocup.Segment(
threat.pos, threat.best_shot_window.segment.center())
else:
shot_line = robocup.Segment(threat.pos, robocup.Point(0, 0))
# debug output
if self.debug:
for handler in threat.assigned_handlers:
# handler.robot.add_text("Marking: " + str(threat.source), constants.Colors.White, "Defense")
main.system_state().draw_circle(handler.move_target, 0.02,
constants.Colors.Blue,
"Defense")
# draw some debug stuff
if threat.best_shot_window != None:
# draw shot triangle
pts = [
threat.pos,
threat.best_shot_window.segment.get_pt(0),
threat.best_shot_window.segment.get_pt(1)
]
shot_color = (255, 0, 0, 150) # translucent red
main.system_state().draw_polygon(pts, shot_color,
"Defense")
main.system_state().draw_segment(
threat.best_shot_window.segment, constants.Colors.Red,
"Defense")
self.win_eval.excluded_robots.clear()
_, best_window = self.win_eval.eval_pt_to_our_goal(
threat.pos)
if best_window is not None:
chance = best_window.shot_success
else:
chance = 0.0
main.system_state().draw_text(
"Shot: " + str(int(threat.shot_chance * 100.0)) +
"% / " + str(int(chance * 100)) + "%",
shot_line.center(), constants.Colors.White, "Defense")
# draw pass lines
if idx > 0:
pass_line = robocup.Segment(main.ball().pos, threat.pos)
main.system_state().draw_line(pass_line,
constants.Colors.Red,
"Defense")
main.system_state().draw_text(
"Pass: "******"%",
pass_line.center(), constants.Colors.White, "Defense")
def execute_running(self):
their_robots = main.their_robots()
mark_one = self.subbehavior_with_name('mark_one')
mark_two = self.subbehavior_with_name('mark_two')
centerCircle = robocup.Circle(constants.Field.CenterPoint,
constants.Field.CenterRadius)
# Don't select robots that are
# 1. Not on our side of the field
# 2. behind or inside the goal circle
mark_robot_right = list(filter(
lambda robot: (robot.pos.x >= 0 and robot.pos.y < constants.Field.Length * TheirKickoff.FieldRatio and constants.Field.FieldRect.contains_point(robot.pos) and not centerCircle.contains_point(robot.pos)),
their_robots))
# Don't select robots that are
# 1. Not on our side of the field
# 2. behind or inside the goal circle
# 3. Not the robot selected before
mark_robot_left = list(filter(
lambda robot: (robot.pos.x <= 0 and robot.pos.y < constants.Field.Length * TheirKickoff.FieldRatio and constants.Field.FieldRect.contains_point(robot.pos) and not centerCircle.contains_point(robot.pos) and robot != mark_one.mark_robot),
their_robots))
# Special cases
if len(mark_robot_left) + len(mark_robot_right) == 0:
# Can't do anything
mark_robot_left = None
mark_robot_right = None
elif len(mark_robot_left) + len(mark_robot_right) == 1:
if len(mark_robot_left) == 1:
mark_robot_right = mark_robot_left[0]
mark_robot_left = None
else:
mark_robot_right = mark_robot_right[0]
mark_robot_left = None
elif len(mark_robot_left) == 0:
mark_robot_right = mark_robot_right
mark_robot_left = mark_robot_right
elif len(mark_robot_right) == 0:
mark_robot_right = mark_robot_left
mark_robot_left = mark_robot_left
# Else, everything can proceed as normal (pick best one from each side)
# Make every element a list to normalize for the next step
if type(mark_robot_right) is not list and mark_robot_right is not None:
mark_robot_right = [mark_robot_right]
if type(mark_robot_left) is not list and mark_robot_left is not None:
mark_robot_left = [mark_robot_left]
# Select best robot from candidate lists
selected = None
if mark_robot_right is not None:
mark_robot_right = min(mark_robot_right,
key=lambda robot: robot.pos.y).pos
selected = robocup.Point(mark_robot_right)
else:
mark_robot_right = robocup.Point(TheirKickoff.DefaultDist,
constants.Field.Length / 2)
# Set x and y seperately as we want a constant y value (just behind the kick off line)
mark_robot_right.y = min(
constants.Field.Length / 2 - TheirKickoff.LineBuffer,
mark_robot_right.y)
mark_robot_right.x = self.absmin(mark_robot_right.x,
TheirKickoff.DefaultDist)
mark_one.mark_point = mark_robot_right
# Do the same thing as above on the left robot.
if mark_robot_left is not None:
# Don't mark the same robot twice
mark_robot_left = filter(
lambda x: True if selected is None else not x.pos.nearly_equals(selected),
mark_robot_left)
mark_robot_left = min(mark_robot_left,
key=lambda robot: robot.pos.y).pos
else:
mark_robot_left = robocup.Point(-TheirKickoff.DefaultDist,
constants.Field.Length / 2)
mark_robot_left.y = min(
constants.Field.Length / 2 - TheirKickoff.LineBuffer,
mark_robot_left.y)
mark_robot_left.x = self.absmin(mark_robot_left.x,
TheirKickoff.DefaultDist)
mark_two.mark_point = mark_robot_left
def get_threat_list(self, unused_threat_handlers):
# List of (position, score, Robot/None)
threats = []
potential_threats = main.their_robots()
# find the primary threat
# if the ball is not moving OR it's moving towards our goal, it's the primary threat
# if it's moving, but not towards our goal, the primary threat is the robot on their team most likely to catch it
if (main.ball().vel.mag() > 0.4):
if evaluation.ball.is_moving_towards_our_goal():
# Add tuple of pos and score
threats.append((main.ball().pos, 1, None))
else:
# Get all potential receivers
potential_receivers = []
for opp in potential_threats:
if self.estimate_potential_recievers_score(opp) == 1:
potential_receivers.append((opp.pos, 1, opp))
if len(potential_receivers) > 0:
# Add best receiver to threats
# TODO Calc shot chance
best_tuple = min(potential_receivers,
key=lambda rcrv_tuple: rcrv_tuple[1])
threats.append((best_tuple[0], .81, best_tuple[2]))
else:
# Just deal with ball if no recievers
threats.append((main.ball().pos, .9, None))
else:
# Assume opp is dribbling ball
if not constants.Field.OurGoalZoneShape.contains_point(
main.ball().pos):
# TODO: Calc shot chance
threats.append((main.ball().pos, 1, None))
# if there are threats, check pass and shot chances
# If the first item is not a ball, it is most likely a pass
if len(threats) > 0 and threats[0][0] != main.ball().pos:
for opp in potential_threats:
# Exclude robots that have been assigned already
excluded_bots = []
for r in map(lambda bhvr: bhvr.robot, unused_threat_handlers):
excluded_bots.append(r)
threats.append(
(opp.pos, self.estimate_risk_score(opp,
excluded_bots), opp))
else:
for opp in potential_threats:
# Exclude all robots
self.kick_eval.excluded_robots.clear()
self.kick_eval.add_excluded_robot(opp)
for r in main.our_robots():
self.kick_eval.add_excluded_robot(r)
point, shotChance = self.kick_eval.eval_pt_to_our_goal(opp.pos)
# Note: 0.5 is a bullshit value
threats.append((opp.pos, 0.5 * shotChance, opp))
# Prevent threats from being below our goal line (causes incorrect pos)
def _adjust_pt(threat):
pt = threat[0]
pt.y = max(pt.y, 0.1)
return (pt, ) + threat[1:]
threats = list(map(_adjust_pt, threats))
return threats
def find_gap(target_pos=constants.Field.TheirGoalSegment.center(), max_shooting_angle=60, robot_offset=8, dist_from_point=.75):
if (not main.ball().valid):
return target_pos
# Find the hole in the defenders to kick at
# The limit is 20 cm so any point past it should be defenders right there
win_eval = robocup.WindowEvaluator(main.system_state())
# 500 cm min circle distance plus the robot width
test_distance = dist_from_point + constants.Robot.Radius
# +- max offset to dodge ball
max_angle = max_shooting_angle * constants.DegreesToRadians
# How much left and right of a robot to give
# Dont make this too big or it will always go far to the right or left of the robots
robot_angle_offset = robot_offset * constants.DegreesToRadians
zero_point = robocup.Point(0, 0)
# Limit the angle so as we get closer, we dont miss the goal completely as much
goal_vector = target_pos - main.ball().pos
max_length_vector = robocup.Point(constants.Field.Length, constants.Field.Width)
goal_limit = (goal_vector.mag() / max_length_vector.mag()) * max_angle
# Limit on one side so we dont directly kick out of bounds
# Add in the angle from the sideline to the target
field_limit = (1 - abs(main.ball().pos.x) / (constants.Field.Width / 2)) * max_angle
field_limit = field_limit + goal_vector.angle_between(robocup.Point(0, 1))
# Limit the angle based on the opponent robots to try and always minimize the
left_robot_limit = 0
right_robot_limit = 0
for robot in main.their_robots():
ball_to_bot = robot.pos - main.ball().pos
# Add an extra radius as wiggle room
# kick eval already deals with the wiggle room so it isn't needed there
if (ball_to_bot.mag() <= test_distance + constants.Robot.Radius):
angle = goal_vector.angle_between(ball_to_bot)
# Try and rotate onto the goal vector
# if we actually do, then the robot is to the right of the ball vector
ball_to_bot.rotate(zero_point, angle)
if (ball_to_bot.angle_between(goal_vector) < 0.01):
right_robot_limit = max(right_robot_limit, angle + robot_angle_offset)
else:
left_robot_limit = max(left_robot_limit, angle + robot_angle_offset)
else:
win_eval.add_excluded_robot(robot)
# Angle limit on each side of the bot->goal vector
left_angle = max_angle
right_angle = max_angle
# Make sure we limit the correct side due to the field
if main.ball().pos.x < 0:
left_angle = min(left_angle, field_limit)
else:
right_angle = min(right_angle, field_limit)
# Limit due to goal
left_angle = min(left_angle, goal_limit)
right_angle = min(right_angle, goal_limit)
# Limit to just over the robots
if (left_robot_limit is not 0):
left_angle = min(left_angle, left_robot_limit)
if (right_robot_limit is not 0):
right_angle = min(right_angle, right_robot_limit)
# Get the angle that we need to rotate the target angle behind the defenders
# since kick eval doesn't support a nonsymmetric angle around a target
rotate_target_angle = (left_angle + -right_angle)/2
target_width = (left_angle + right_angle)
target_point = goal_vector.normalized() * test_distance
target_point.rotate(zero_point, rotate_target_angle)
windows, window = win_eval.eval_pt_to_pt(main.ball().pos, target_point + main.ball().pos, target_width)
# Test draw points
target_point.rotate(zero_point, target_width/2)
p1 = target_point + main.ball().pos
target_point.rotate(zero_point, -target_width)
p2 = target_point + main.ball().pos
p3 = main.ball().pos
main.system_state().draw_polygon([p1, p2, p3], (0, 0, 255), "Free Kick search zone")
is_opponent_blocking = False
for robot in main.their_robots():
if (goal_vector.dist_to(robot.pos) < constants.Robot.Radius and
(main.ball().pos - robot.pos).mag() < test_distance):
is_opponent_blocking = True
# Vector from ball position to the goal
ideal_shot = (target_pos - main.ball().pos).normalized()
# If on our side of the field and there are enemy robots around us,
# prioritize passing forward vs passing towards their goal
# Would have to change this if we are not aiming for their goal
if main.ball().pos.y < constants.Field.Length / 2 and len(windows) > 1:
ideal_shot = robocup.Point(0, 1)
main.system_state().draw_line(robocup.Line(main.ball().pos, target_pos), (0, 255, 0), "Target Point")
# Weights for determining best shot
k1 = 1.5 # Weight of closeness to ideal shot
k2 = 1 # Weight of shot chance
# Iterate through all possible windows to find the best possible shot
if windows:
best_shot = window.segment.center()
best_weight = 0
for wind in windows:
pos_to_wind = (wind.segment.center() - main.ball().pos).normalized()
dot_prod = pos_to_wind.dot(ideal_shot)
weight = k1 * dot_prod + k2 * wind.shot_success
if weight > best_weight:
best_weight = weight
best_shot = wind.segment.center()
main.system_state().draw_line(robocup.Line(main.ball().pos, best_shot), (255, 255, 0), "Target Shot")
best_shot = robocup.Point(0,1) + main.ball().pos
return best_shot
else:
return constants.Field.TheirGoalSegment.center()
def execute_running(self):
super().execute_running()
striker = self.subbehavior_with_name('striker')
support1 = self.subbehavior_with_name('support1')
support2 = self.subbehavior_with_name('support2')
supports = [support1, support2]
# project ball location a bit into the future
ball_proj = evaluation.ball.predict(main.ball().pos,
main.ball().vel,
t=0.75)
# find closest opponent to striker
closest_dist_to_striker, closest_opp_to_striker = float("inf"), None
if striker.robot != None:
for opp in main.their_robots():
d = opp.pos.dist_to(striker.robot.pos)
if d < closest_dist_to_striker:
closest_dist_to_striker, closest_opp_to_striker = d, opp
striker_engaged = striker.robot != None and closest_dist_to_striker < Basic122.SupportBackoffThresh
# pick out which opponents our defenders should 'mark'
# TODO: explain
nrOppClose = 0
bestOpp1, bestOpp2 = None, None
bestDistSq1, bestDistSq2 = float("inf"), float("inf")
for opp in main.their_robots():
# use dist from goal rather than just y-coord to handle corners better
oppDistSq = opp.pos.magsq()
if not (striker_engaged and opp == closest_opp_to_striker):
if oppDistSq < bestDistSq1:
bestDistSq2, bestOpp2 = bestDistSq1, bestOpp1
bestDistSq1, bestOpp1 = oppDistSq, opp
elif oppDistSq < bestDistSq2:
bestDistSq2, bestOpp2 = oppDistSq, opp
if oppDistSq < constants.Field.Length**2 / 4.0:
nrOppClose += 1
# handle the case of having no good robots to mark
if bestOpp1 == None and support1.robot != None:
support1.mark_robot = None
support1.robot.add_text("No mark target", (255, 255, 255),
"RobotText")
if striker.robot != None:
support1.robot.add_text("Static Support", (255, 255, 255),
"RobotText")
support_goal = striker.robot.pos
support_goal.x *= -1.0
if abs(support_goal.x) < 0.2:
support_goal.x = -1.0 if support_goal.x < 0 else 1.0
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
support_goal.y = max(support_goal.y *
Basic122.OffenseSupportRatio, 0.3)
else:
support_goal.y = max(support_goal.y *
Basic122.DefenseSupportRatio, 0.3)
support1.robot.move_to(support_goal)
support1.robot.face(ball_proj)
# sit around and do jack shit...
# TODO: make it do something useful
if bestOpp2 == None and support2.robot != None:
support2.mark_robot = None
support2.robot.add_text("No mark target", (255, 255, 255),
"RobotText")
if striker.robot != None:
support2.robot.add_text("Static Support", (255, 255, 255),
"RobotText")
support_goal = striker.robot.pos
support_goal.x *= -1.0
if abs(support_goal.x) < 0.2:
support_goal.x = -1.0 if support_goal.x < 0 else 1.0
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
support_goal.y = max(support_goal.y *
Basic122.OffenseSupportRatio, 0.3)
else:
support_goal.y = max(support_goal.y *
Basic122.DefenseSupportRatio, 0.3)
support2.robot.move_to(support_goal)
support2.robot.face(ball_proj)
# reassign support robots' mark targets based on dist sq and hysteresis coeff
new_dists = [bestDistSq1, bestDistSq2]
new_bots = [bestOpp1, bestOpp2]
for i in range(2):
support = supports[i]
if new_bots[i] != None:
cur_dist_sq = (
support.mark_robot.pos -
ball_proj).magsq() if support.mark_robot else float("inf")
if new_dists[i] < cur_dist_sq * Basic122.MarkHysteresisCoeff:
support.mark_robot = new_bots[i]
# if the supports are farther from the ball, they can mark further away
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
for support in supports:
support.ratio = Basic122.OffenseSupportRatio
else:
for support in supports:
support.ratio = Basic122.DefenseSupportRatio
# raise NotImplementedError("Make support robots avoid the shot channel")
# FROM C++:
# Polygon shot_obs;
# shot_obs.vertices.push_back(Geometry2d::Point(Field_GoalWidth / 2, Field_Length));
# shot_obs.vertices.push_back(Geometry2d::Point(-Field_GoalWidth / 2, Field_Length));
# shot_obs.vertices.push_back(ballProj);
# TODO: this would be a good place for a "keep-trying container behavior"
# make the kicker try again if it already kicked
if not striker.is_in_state(behavior.Behavior.State.running):
striker.restart()
def on_enter_running(self):
b = skills.goalside_mark.Goalside_Mark()
b.mark_robot = main.their_robots()[0]
self.add_subbehavior(b, name='goalside mark', required=True)
def find_gap(target_pos=constants.Field.TheirGoalSegment.center(), max_shooting_angle=60, robot_offset=8, dist_from_point=.75):
if (not main.ball().valid):
return target_pos
# Find the hole in the defenders to kick at
# The limit is 20 cm so any point past it should be defenders right there
win_eval = robocup.WindowEvaluator(main.system_state())
# 500 cm min circle distance plus the robot width
test_distance = dist_from_point + constants.Robot.Radius
# +- max offset to dodge ball
max_angle = max_shooting_angle * constants.DegreesToRadians
# How much left and right of a robot to give
# Dont make this too big or it will always go far to the right or left of the robots
robot_angle_offset = robot_offset * constants.DegreesToRadians
zero_point = robocup.Point(0, 0)
# Limit the angle so as we get closer, we dont miss the goal completely as much
goal_vector = target_pos - main.ball().pos
max_length_vector = robocup.Point(constants.Field.Length, constants.Field.Width)
goal_limit = (goal_vector.mag() / max_length_vector.mag()) * max_angle
# Limit on one side so we dont directly kick out of bounds
# Add in the angle from the sideline to the target
field_limit = (1 - abs(main.ball().pos.x) / (constants.Field.Width / 2)) * max_angle
field_limit = field_limit + goal_vector.angle_between(robocup.Point(0, 1))
# Limit the angle based on the opponent robots to try and always minimize the
left_robot_limit = 0
right_robot_limit = 0
for robot in main.their_robots():
ball_to_bot = robot.pos - main.ball().pos
# Add an extra radius as wiggle room
# kick eval already deals with the wiggle room so it isn't needed there
if (ball_to_bot.mag() <= test_distance + constants.Robot.Radius):
angle = goal_vector.angle_between(ball_to_bot)
# Try and rotate onto the goal vector
# if we actually do, then the robot is to the right of the ball vector
ball_to_bot.rotate(zero_point, angle)
if (ball_to_bot.angle_between(goal_vector) < 0.01):
right_robot_limit = max(right_robot_limit, angle + robot_angle_offset)
else:
left_robot_limit = max(left_robot_limit, angle + robot_angle_offset)
else:
win_eval.add_excluded_robot(robot)
# Angle limit on each side of the bot->goal vector
left_angle = max_angle
right_angle = max_angle
# Make sure we limit the correct side due to the field
if main.ball().pos.x < 0:
left_angle = min(left_angle, field_limit)
else:
right_angle = min(right_angle, field_limit)
# Limit due to goal
left_angle = min(left_angle, goal_limit)
right_angle = min(right_angle, goal_limit)
# Limit to just over the robots
if (left_robot_limit is not 0):
left_angle = min(left_angle, left_robot_limit)
if (right_robot_limit is not 0):
right_angle = min(right_angle, right_robot_limit)
# Get the angle that we need to rotate the target angle behind the defenders
# since kick eval doesn't support a nonsymmetric angle around a target
rotate_target_angle = (left_angle + -right_angle)/2
target_width = (left_angle + right_angle)
target_point = goal_vector.normalized() * test_distance
target_point.rotate(zero_point, rotate_target_angle)
windows, window = win_eval.eval_pt_to_pt(main.ball().pos, target_point + main.ball().pos, target_width)
# Test draw points
target_point.rotate(zero_point, target_width/2)
p1 = target_point + main.ball().pos
target_point.rotate(zero_point, -target_width)
p2 = target_point + main.ball().pos
p3 = main.ball().pos
main.system_state().draw_polygon([p1, p2, p3], (0, 0, 255), "Free Kick search zone")
is_opponent_blocking = False
for robot in main.their_robots():
if (goal_vector.dist_to(robot.pos) < constants.Robot.Radius and
(main.ball().pos - robot.pos).mag() < test_distance):
is_opponent_blocking = True
# Vector from ball position to the goal
ideal_shot = (target_pos - main.ball().pos).normalized()
# If on our side of the field and there are enemy robots around us,
# prioritize passing forward vs passing towards their goal
# Would have to change this if we are not aiming for their goal
if main.ball().pos.y < constants.Field.Length / 2 and len(windows) > 1:
ideal_shot = robocup.Point(0, 1)
main.system_state().draw_line(robocup.Line(main.ball().pos, target_pos), (0, 255, 0), "Target Point")
# Weights for determining best shot
k1 = 1.5 # Weight of closeness to ideal shot
k2 = 1 # Weight of shot chance
# print("ideal shot:", ideal_shot)
# print("number of windows:", len(windows))
# Iterate through all possible windows to find the best possible shot
if windows:
best_shot = window.segment.center()
best_weight = 0
for wind in windows:
pos_to_wind = (wind.segment.center() - main.ball().pos).normalized()
dot_prod = pos_to_wind.dot(ideal_shot)
# print("weighted dot product:", dot_prod)
# print("weighted shot chance:", wind.shot_success)
weight = k1 * dot_prod + k2 * wind.shot_success
if weight > best_weight:
best_weight = weight
best_shot = wind.segment.center()
main.system_state().draw_line(robocup.Line(main.ball().pos, best_shot), (255, 255, 0), "Target Shot")
best_shot = robocup.Point(0,1) + main.ball().pos
return best_shot
else:
return constants.Field.TheirGoalSegment.center()
def execute_running(self):
striker = self.subbehavior_with_name('striker')
support1 = self.subbehavior_with_name('support1')
support2 = self.subbehavior_with_name('support2')
supports = [support1, support2]
# project ball location a bit into the future
ball_proj = evaluation.ball.predict(main.ball().pos, main.ball().vel, t=0.75)
# find closest opponent to striker
closest_dist_to_striker, closest_opp_to_striker = float("inf"), None
if striker.robot != None:
for opp in main.their_robots():
d = opp.pos.dist_to(striker.robot.pos)
if d < closest_dist_to_striker:
closest_dist_to_striker, closest_opp_to_striker = d, opp
striker_engaged = striker.robot != None and closest_dist_to_striker < Basic122.SupportBackoffThresh
# pick out which opponents our defenders should 'mark'
# TODO: explain
nrOppClose = 0
bestOpp1, bestOpp2 = None, None
bestDistSq1, bestDistSq2 = float("inf"), float("inf")
for opp in main.their_robots():
# use dist from goal rather than just y-coord to handle corners better
oppDistSq = opp.pos.magsq()
if not (striker_engaged and opp == closest_opp_to_striker):
if oppDistSq < bestDistSq1:
bestDistSq2, bestOpp2 = bestDistSq1, bestOpp1
bestDistSq1, bestOpp1 = oppDistSq, opp
elif oppDistSq < bestDistSq2:
bestDistSq2, bestOpp2 = oppDistSq, opp
if oppDistSq < constants.Field.Length**2 / 4.0:
nrOppClose += 1
# handle the case of having no good robots to mark
if bestOpp1 == None and support1.robot != None:
support1.mark_robot = None
support1.robot.add_text("No mark target", (255,255,255), "RobotText")
if striker.robot != None:
support1.robot.add_text("Static Support", (255,255,255), "RobotText")
support_goal = striker.robot.pos
support_goal.x *= -1.0
if abs(support_goal.x) < 0.2:
support_goal.x = -1.0 if support_goal.x < 0 else 1.0
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
support_goal.y = max(support_goal.y * Basic122.OffenseSupportRatio, 0.3)
else:
support_goal.y = max(support_goal.y * Basic122.DefenseSupportRatio, 0.3)
support1.robot.move_to(support_goal)
support1.robot.face(ball_proj)
# sit around and do jack shit...
# TODO: make it do something useful
if bestOpp2 == None and support2.robot != None:
support2.mark_robot = None
support2.robot.add_text("No mark target", (255,255,255), "RobotText")
if striker.robot != None:
support2.robot.add_text("Static Support", (255,255,255), "RobotText")
support_goal = striker.robot.pos
support_goal.x *= -1.0
if abs(support_goal.x) < 0.2:
support_goal.x = -1.0 if support_goal.x < 0 else 1.0
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
support_goal.y = max(support_goal.y * Basic122.OffenseSupportRatio, 0.3)
else:
support_goal.y = max(support_goal.y * Basic122.DefenseSupportRatio, 0.3)
support2.robot.move_to(support_goal)
support2.robot.face(ball_proj)
# reassign support robots' mark targets based on dist sq and hysteresis coeff
new_dists = [bestDistSq1, bestDistSq2]
new_bots = [bestOpp1, bestOpp2]
for i in range(2):
support = supports[i]
if new_bots[i] != None:
cur_dist_sq = (support.mark_robot.pos - ball_proj).magsq() if support.mark_robot else float("inf")
if new_dists[i] < cur_dist_sq * Basic122.MarkHysteresisCoeff:
support.mark_robot = new_bots[i]
# if the supports are farther from the ball, they can mark further away
if ball_proj.y > constants.Field.Length / 2.0 and nrOppClose > 0:
for support in supports: support.ratio = Basic122.OffenseSupportRatio
else:
for support in supports: support.ratio = Basic122.DefenseSupportRatio
# keep support robots away from the striker
if striker.robot != None:
for supp in [support1, support2]:
if supp.robot != None:
supp.robot.set_avoid_teammate_radius(striker.robot.shell_id(), Basic122.SupportAvoidTeammateRadius)
# raise NotImplementedError("Make support robots avoid the shot channel")
# FROM C++:
# Polygon shot_obs;
# shot_obs.vertices.push_back(Geometry2d::Point(Field_GoalWidth / 2, Field_Length));
# shot_obs.vertices.push_back(Geometry2d::Point(-Field_GoalWidth / 2, Field_Length));
# shot_obs.vertices.push_back(ballProj);
# TODO: this would be a good place for a "keep-trying container behavior"
# make the kicker try again if it already kicked
if not striker.is_in_state(behavior.Behavior.State.running):
striker.restart()
def has_ball(self):
closeIndicator = False
for r in main.their_robots():
closeIndicator = closeIndicator or (
r.pos - main.ball().pos).mag() < self.ball_lost_distance
return closeIndicator