def plan(self, observation):
envs = [
preprocess_env(self.true_env, preprocessors)
for preprocessors in self.config["models"]
]
self.env = JointEnv(envs)
return super().plan(observation)
def display(cls, agent, agent_surface, sim_surface):
import pygame
horizon = 2
robust_env = preprocess_env(agent.env,
agent.config["env_preprocessors"])
for vehicle in robust_env.road.vehicles:
vehicle.COLLISIONS_ENABLED = False
plan = agent.get_plan()
if plan:
plan = plan[1:] # First action has already been performed
plan = plan[:horizon] + (horizon - len(plan)) * [1]
for action in plan:
robust_env.step(action)
for vehicle in robust_env.road.vehicles:
if not hasattr(vehicle, 'interval_trajectory'):
continue
min_traj = [o.position[0] for o in vehicle.interval_trajectory]
max_traj = [o.position[1] for o in vehicle.interval_trajectory]
uncertainty_surface = pygame.Surface(sim_surface.get_size(),
pygame.SRCALPHA, 32)
cls.display_uncertainty(min_traj, max_traj, uncertainty_surface,
sim_surface, cls.UNCERTAINTY_TIME_COLORMAP)
cls.display_trajectory(vehicle.trajectory, uncertainty_surface,
sim_surface, cls.MODEL_TRAJ_COLOR)
sim_surface.blit(uncertainty_surface, (0, 0))
if agent_surface and hasattr(agent, "sub_agent"):
TreeGraphics.display(agent.sub_agent, agent_surface)
def plan(self, observation):
"""
Plan an optimal sequence of actions.
Start by updating the previously found tree with the last action performed.
:param observation: the current state
:return: the list of actions
"""
self.planner.step(self.previous_action)
env = preprocess_env(self.env, self.config["env_preprocessors"])
actions = self.planner.plan(state=env, observation=observation)
self.previous_action = actions[0]
return actions
def display(cls, agent, agent_surface, sim_surface):
import pygame
horizon = 2
plan = agent.planner.get_plan()
for env in [preprocess_env(agent.true_env, preprocessors) for preprocessors in agent.config["models"]]:
for vehicle in env.road.vehicles:
vehicle.trajectory = []
for action in plan[:horizon] + (horizon - len(plan)) * [1]:
env.step(action)
for vehicle in env.road.vehicles:
if vehicle is env.vehicle:
continue
uncertainty_surface = pygame.Surface(sim_surface.get_size(), pygame.SRCALPHA, 32)
IntervalRobustPlannerGraphics.display_trajectory(vehicle.trajectory, uncertainty_surface, sim_surface,
IntervalRobustPlannerGraphics.MODEL_TRAJ_COLOR)
sim_surface.blit(uncertainty_surface, (0, 0))
TreeGraphics.display(agent, agent_surface)
def display(cls, agent, agent_surface, sim_surface):
plan = agent.planner.get_plan()
for env in [preprocess_env(agent.true_env, preprocessors) for preprocessors in agent.config["models"]]:
IntervalRobustPlannerGraphics.display_uncertainty(env, plan, sim_surface, trajectory=False)
# for vehicle in env.road.vehicles:
# if hasattr(vehicle, ):
# IntervalRobustPlannerGraphics.display(vehicle)
# for vehicle in env.road.vehicles:
# vehicle.trajectory = []
# for action in plan[:horizon] + (horizon - len(plan)) * [1]:
# env.step(action)
# for vehicle in env.road.vehicles:
# if vehicle is env.vehicle:
# continue
# uncertainty_surface = pygame.Surface(sim_surface.get_size(), pygame.SRCALPHA, 32)
# IntervalRobustPlannerGraphics.display_trajectory(vehicle.trajectory, uncertainty_surface, sim_surface,
# IntervalRobustPlannerGraphics.MODEL_TRAJ_COLOR)
# sim_surface.blit(uncertainty_surface, (0, 0))
TreeGraphics.display(agent, agent_surface)
def plan(self, observation):
"""
Plan an optimal sequence of actions.
Start by updating the previously found tree with the last action performed.
:param observation: the current state
:return: the list of actions
"""
self.steps += 1
replanning_required = self.step(self.previous_actions)
if replanning_required:
env = preprocess_env(self.env, self.config["env_preprocessors"])
actions = self.planner.plan(state=env, observation=observation)
else:
actions = self.previous_actions[1:]
self.write_tree()
self.previous_actions = actions
return actions
def plan(self, observation):
self.sub_agent.env = preprocess_env(self.env, self.config["env_preprocessors"])
return self.sub_agent.plan(observation)
def plan(self, observation):
envs = [preprocess_env(self.true_env, preprocessors) for preprocessors in self.config["models"]]
self.env = JointEnv(envs)
return super(DiscreteRobustPlannerAgent, self).plan(observation)
def display(cls, agent, agent_surface, sim_surface):
robust_env = preprocess_env(agent.env, agent.config["env_preprocessors"])
cls.display_uncertainty(robust_env, plan=agent.get_plan(), surface=sim_surface)
if agent_surface and hasattr(agent, "sub_agent"):
TreeGraphics.display(agent.sub_agent, agent_surface)
