def dummyLoop():
global maxSpeed, shotAngle, wallClose, dead, previousScore
global turnedLeft, turnedRight, thrusted, shot
#Release keys
DataMinerBD.tthrustDummy(0)
DataMinerBD.tturnLeftDummy(0)
DataMinerBD.tturnRightDummy(0)
ai.setTurnSpeed(45)
#Set variables"""
heading = int(ai.selfHeadingDeg())
tracking = int(ai.selfTrackingDeg())
trackWall = ai.wallFeeler(500, tracking)
trackLWall = ai.wallFeeler(500, tracking + 3)
trackRWall = ai.wallFeeler(500, tracking - 3)
frontWall = ai.wallFeeler(500, heading)
flWall = ai.wallFeeler(500, heading + 10)
frWall = ai.wallFeeler(500, heading - 10)
leftWall = ai.wallFeeler(500, heading + 90)
llWall = ai.wallFeeler(500, heading + 100)
rlWall = ai.wallFeeler(500, heading + 80)
rightWall = ai.wallFeeler(500, heading - 90)
lrWall = ai.wallFeeler(500, heading - 80)
rrWall = ai.wallFeeler(500, heading - 100)
trackWall = ai.wallFeeler(500, tracking)
backWall = ai.wallFeeler(500, heading - 180)
backLeftWall = ai.wallFeeler(500, heading - 190)
backRightWall = ai.wallFeeler(500, heading - 170)
speed = ai.selfSpeed()
closest = min(frontWall, leftWall, rightWall, backWall, flWall,
frWall)
def closestWall(x): #Find the closest Wall
return {
frontWall: 1,
leftWall: 2,
rightWall: 3,
backWall: 4,
flWall: 5,
frWall: 6,
}[x]
wallNum = closestWall(closest)
#Code for finding the angle to the closest ship
targetX, targetY = ai.screenEnemyX(0), ai.screenEnemyY(0)
#baseString = "["+str(flWall/500)+","+str(frontWall/500)+","+str(frWall/500) + "," + str(backLeftWall/500) + "," + str(backWall/500) + "," + str(backRightWall/500) + ","+str(leftWall/500)+","+str(rightWall/500)+","+str(trackLWall/500) + "," + str(trackWall/500) + ","+str(trackRWall/500) + "," + str(speed/10)
calcDir = -1
if targetX - ai.selfX() != 0:
calcDir = (math.degrees(
math.atan2((targetY - ai.selfY()),
(targetX - ai.selfX()))) + 360) % 360
crashWall = min(
trackWall, trackLWall, trackRWall
) #The wall we are likely to crash into if we continue on our current course
#Rules for turning
if crashWall > wallClose * speed and closest > 25 and targetX != -1: #If we are far enough away from a predicted crash and no closer than 25 pixels to a wall we can try and aim and kill them
diff = (calcDir - heading)
#if ai.shotAlert(0) > -1 and ai.shotAlert(0) < 35: #If we are about to get shot
# tturnRight(1) #Screw aiming and turn right and thrust
# tthrust(1) #This is arguably a horrible strategy because our sideways profile is much larger, but it's required for the grade
if diff >= 0:
if diff >= 180:
DataMinerBD.tturnRightDummy(
1) #If the target is to our right- turn right
else:
DataMinerBD.tturnLeftDummy(
1) #If the target is to our left - turn left
else:
if diff > -180:
DataMinerBD.tturnRightDummy(
1) #If the target is to our right - turn right
else:
DataMinerBD.tturnLeftDummy(
1) #If the target is to our left - turn left
else: #Rules for avoiding death
# if crashWall/ai.selfSpeed() > ai.closestShot() :
if wallNum == 1 or wallNum == 5 or wallNum == 6: #Front Wall is Closest (Turn Away From It)
DataMinerBD.tturnLeftDummy(1)
elif wallNum == 2: # Left Wall is Closest (Turn Away From It)
DataMinerBD.tturnRightDummy(1)
elif wallNum == 3: #Right Wall is Closest (Turn Away From It)
DataMinerBD.tturnLeftDummy(1)
else: #Back Wall is closest- turn so that we are facing directly away from it
if backLeftWall < backRightWall:
DataMinerBD.tturnRightDummy(
1
) #We need to turn right to face more directly away from it
if backLeftWall > backRightWall: # We need to turn left to face more directly away from it
DataMinerBD.tturnLeftDummy(1)
#Rules for thrusting
if speed < maxSpeed and frontWall > 100: #If we are moving slowly and we won't ram into anything, accelerate
DataMinerBD.tthrustDummy(1)
elif trackWall < 200 and (
ai.angleDiff(heading, tracking) > 120
): #If we are getting close to a wall, and we can thrust away from it, do so
DataMinerBD.tthrustDummy(1)
elif backWall < 20: #If there is a wall very close behind us, get away from it
DataMinerBD.tthrustDummy(1)
if abs(
calcDir - heading
) < shotAngle and calcDir != -1: #If we are close to the current proper trajectory for a shot then fire
DataMinerBD.tshootDummy()
previousScore = ai.selfScore()
def AI_loop():
global count_frame, loop, boolean, score, population_size, chromosome_size, population, mutation_prob, crossover_prob, fitness_list, generation, generation_size, first_time, done_learning
#Release keys
ai.thrust(0)
ai.turnLeft(0)
ai.turnRight(0)
## Get A Chromosome in the Population -- Eventually Will go through each individual in the population ##
current_chromosome = population[loop]
## Transform Each Gene insisde A Single Selected Chromosome. 0s & 1s Are Turned Into Intergers For Fuzzy Sets to understand ##
## Each Value obtained is used to calculate the risk of each 45 degree around the agent ##
## Each value has its own "jump" variable which refers to the distance from each possible points/values ##
## The start and end represents the possible start point and end point for each variable and they depend on
## what the variiable is. It is to ensure a viable fuzzy set and fuzzy functions that these restrictions are applied. ##
frontAlert = current_chromosome[0:5]
frontAlertValue = transform(frontAlert, 25)
backAlert = current_chromosome[5:9]
backAlertValue = transform(backAlert, 25)
speedAlert = current_chromosome[9:13] #4 bits
speedAlertValue = transform(speedAlert, 1) #1 jumps per value
EnemyAlert = current_chromosome[13:18] #5 bits
EnemyAlertValue = transform(EnemyAlert, 50) #50 jumps per value
TrackSlowAlert = current_chromosome[18:22] #4 bits
TrackSlowAlertValue = transform(TrackSlowAlert, 25) #25 jumps per value
TrackFastAlert = current_chromosome[22:26] #4 bits
TrackFastAlertValue = transform(TrackFastAlert, 25) #25 jumps per value
BulletAlert = current_chromosome[26:32] #4 bits
BulletAlertValue = transform(BulletAlert, 15) #15 jumps per value
## Get values of variables for Wall Feelers, Head & Tracking ##
heading = int(ai.selfHeadingDeg())
tracking = int(ai.selfTrackingDeg())
frontWall = ai.wallFeeler(500,heading)
left45Wall = ai.wallFeeler(500,heading+45)
right45Wall = ai.wallFeeler(500,heading-45)
left90Wall = ai.wallFeeler(500,heading+90)
right90Wall = ai.wallFeeler(500,heading-90)
left135Wall = ai.wallFeeler(500,heading+135)
right135Wall = ai.wallFeeler(500,heading-135)
backWall = ai.wallFeeler(500,heading-180)
trackWall = ai.wallFeeler(500,tracking)
####### Getters Variable Regarding Important Information About Enemies ########
##Find the closest ennemy##
enemy = ai.lockClose()
## Get the lockheadingdeg of enemy ##
head = ai.lockHeadingDeg()
## Get the dstance from enemy ##
enemyDist = ai.selfLockDist()
## If the Enemy is Dead ##
if(ai.selfAlive() == 0 and boolean == False):
## Calculate Fitness Current Population ##
score_previous = score
score_current = ai.selfScore()
fitness_value = fitness(population, count_frame, score_previous, score_current)
fitness_list.append(fitness_value)
## If it went through the whole population and ready to move to next generation ##
if((loop+1) == population_size):
## Output the fitness of population to allow user to see if learning is happening ##
print("Generation:", generation)
print("Agent Fitness:")
print(fitness_list)
print("Average Fitness:", statistics.mean(fitness_list))
print("Best Fitness:", max(fitness_list))
## Finding the optimal chromosome to output it in data file ##
string_maxChromosome = ""
for chrom_max in range(chromosome_size):
string_maxChromosome = string_maxChromosome + str(population[fitness_list.index(max(fitness_list))][chrom_max])
## Formatting entire population in a big string to register it in excel file##
string_population = ""
for pop in range(population_size):
for pop_chrom in range(chromosome_size):
string_population = string_population + str(population[pop][pop_chrom])
if(pop != (population_size-1)):
string_population = string_population + ","
## Formatting entire population's fitness in a big string to register it in excel file##
string_fitness = ""
for fit in range(len(fitness_list)):
string_fitness = string_fitness + str(fitness_list[fit])
if(fit != (len(fitness_list)-1)):
string_fitness = string_fitness + ","
## Output Data into Excel File ##
titles = ["Generation", "Average Fitness", "Best Fitness","Population Size", "Chromosome Size", "Crossover Probability", "Mutation Probability", "Best Chromosome", "Entire Population Chromosome", "Entire Population Fitness"]
data = [generation, statistics.mean(fitness_list), max(fitness_list), population_size, chromosome_size, crossover_prob, mutation_prob, string_maxChromosome, string_population, string_fitness]
first_time = Save_Data("Tiger_Training_Data.xls", 0, titles, data, first_time)
## Select Next Generation -- Apply Crossover & Mutation ##
new_population = select(population, fitness_list)
new_population = crossover(new_population, chromosome_size, population_size, crossover_prob)
new_population = mutate(new_population, chromosome_size, mutation_prob)
population = new_population
loop = 0
count_frame = 0
generation += 1
fitness_list.clear()
### DONE -- QUIT ###
if (generation == generation_size):
quitAI()
## Move to the next individual in population ##
else:
loop += 1
count_frame = 0
boolean = True
else:
## The agent is Alive ##
if(ai.selfAlive() == 1):
### Turning Rules ###
if frontWall <= frontAlertValue and (left45Wall < right45Wall) and ai.selfSpeed() > speedAlertValue:
ai.turnRight(1)
elif frontWall <= frontAlertValue and (left45Wall > right45Wall) and ai.selfSpeed() > speedAlertValue:
ai.turnLeft(1)
elif left90Wall <= frontAlertValue and ai.selfSpeed() > speedAlertValue:
ai.turnRight(1)
elif right90Wall <= frontAlertValue and ai.selfSpeed() > speedAlertValue:
ai.turnLeft(1)
### Thrust commands ####
elif ai.selfSpeed() <= speedAlertValue and (frontWall >= frontAlertValue) and (left45Wall >= frontAlertValue) and (right45Wall >= frontAlertValue) and (right90Wall >= frontAlertValue) and (left90Wall >= frontAlertValue) and (left135Wall >= backAlertValue) and (right135Wall >= backAlertValue) and (backWall >= backAlertValue):
ai.thrust(1)
elif trackWall <= TrackFastAlertValue and ai.selfSpeed() >= speedAlertValue:
ai.thrust(1)
elif trackWall <= TrackSlowAlertValue and ai.selfSpeed() <= speedAlertValue:
ai.thrust(1)
elif backWall <= TrackFastAlertValue and ai.selfSpeed() >= speedAlertValue:
ai.thrust(1)
elif backWall <= TrackSlowAlertValue and ai.selfSpeed() <= speedAlertValue:
ai.thrust(1)
elif left135Wall <= TrackFastAlertValue and ai.selfSpeed() >= speedAlertValue:
ai.thrust(1)
elif left135Wall <= TrackSlowAlertValue and ai.selfSpeed() <= speedAlertValue:
ai.thrust(1)
elif right135Wall <= TrackFastAlertValue and ai.selfSpeed() >= speedAlertValue:
ai.thrust(1)
elif right135Wall <= TrackSlowAlertValue and ai.selfSpeed() <= speedAlertValue:
ai.thrust(1)
##### Bullet Avoidance Commands #####
elif ai.shotAlert(0) >= 0 and ai.shotAlert(0) <= BulletAlertValue:
if ai.angleDiff(heading, ai.shotVelDir(0)) > 0 and ai.selfSpeed() <= speedAlertValue:
ai.turnLeft(1)
ai.thrust(1)
elif ai.angleDiff(heading, ai.shotVelDir(0)) < 0 and ai.selfSpeed() <= speedAlertValue:
ai.turnRight(1)
ai.thrust(1)
elif ai.angleDiff(heading, ai.shotVelDir(0)) > 0 and ai.selfSpeed() > speedAlertValue:
ai.turnLeft(1)
else:
ai.turnRight(1)
##### Shooting Ennemy Commands #####
elif enemyDist <= EnemyAlertValue and heading > (head) and ai.selfSpeed() > speedAlertValue:
ai.turnRight(1)
ai.fireShot()
elif enemyDist <= EnemyAlertValue and heading < (head) and ai.selfSpeed() > speedAlertValue:
ai.turnLeft(1)
ai.fireShot()
elif ai.selfSpeed() < speedAlertValue:
ai.thrust(1)
else:
ai.thrust(0)
count_frame += 3
boolean = False
def AI_loop():
global count_frame, loop, boolean, score, population_size, chromosome_size, population, mutation_prob, crossover_prob, fitness_list, generation, generation_size, first_time, done_learning
#Release keys
ai.thrust(0)
ai.turnLeft(0)
ai.turnRight(0)
## Get A Chromosome in the Population -- Eventually Will go through each individual in the population ##
current_chromosome = population[loop]
## Transform Each Gene insisde A Single Selected Chromosome. 0s & 1s Are Turned Into Intergers For Fuzzy Sets to understand ##
## Each Value obtained is used to calculate the risk of each 45 degree around the agent ##
## Each value has its own "jump" variable which refers to the distance from each possible points/values ##
## The start and end represents the possible start point and end point for each variable and they depend on
## what the variiable is. It is to ensure a viable fuzzy set and fuzzy functions that these restrictions are applied. ##
closingRate_SlowTopAlert = current_chromosome[0:4]
closingRate_SlowTopAlertValue = transform_fuzzy(closingRate_SlowTopAlert,
1, 0, 16)
closingRate_MediumTopLeftAlert = current_chromosome[4:8]
closingRate_MediumTopLeftAlertValue = transform_fuzzy(
closingRate_MediumTopLeftAlert, 1, (closingRate_SlowTopAlertValue + 1),
(closingRate_SlowTopAlertValue + 1) + 16)
closingRate_MediumTopRightAlert = current_chromosome[8:12]
closingRate_MediumTopRightAlertValue = transform_fuzzy(
closingRate_MediumTopRightAlert, 1,
(closingRate_MediumTopLeftAlertValue + 1),
(closingRate_MediumTopLeftAlertValue + 1) + 16)
closingRate_FastTopAlert = current_chromosome[12:16]
closingRate_FastTopAlertValue = transform_fuzzy(
closingRate_FastTopAlert, 1,
(closingRate_MediumTopRightAlertValue + 1),
(closingRate_MediumTopRightAlertValue + 1) + 16)
closingRate_SlowBottomAlert = current_chromosome[16:20]
start = (closingRate_SlowTopAlertValue +
(((closingRate_MediumTopLeftAlertValue -
closingRate_SlowTopAlertValue) // 2) + 1))
end = (start + (1 * (2**(len(closingRate_SlowBottomAlert)))))
closingRate_SlowBottomAlertValue = transform_fuzzy(
closingRate_SlowBottomAlert, 1, start, end)
closingRate_MediumBottomLeftAlert = current_chromosome[20:24]
end = (closingRate_MediumTopLeftAlertValue -
(((closingRate_MediumTopLeftAlertValue -
closingRate_SlowTopAlertValue) // 2) + 1))
start = end - (1 * (2**(len(closingRate_MediumBottomLeftAlert))))
if (end < 0):
end = 0
if (start < 0):
start = 0
jump = (end - start) // (2**(len(closingRate_MediumBottomLeftAlert)))
closingRate_MediumBottomLeftAlertValue = transform_fuzzy(
closingRate_MediumBottomLeftAlert, jump, start, end)
closingRate_MediumBottomRightAlert = current_chromosome[24:28]
start = (closingRate_MediumTopRightAlertValue +
(((closingRate_FastTopAlertValue -
closingRate_MediumTopRightAlertValue) // 2) + 1))
end = start + (1 * (2**(len(closingRate_MediumBottomRightAlert))))
closingRate_MediumBottomRightAlertValue = transform_fuzzy(
closingRate_MediumBottomRightAlert, 1, start, end)
closingRate_FastBottomAlert = current_chromosome[28:32]
end = (closingRate_FastTopAlertValue -
(((closingRate_FastTopAlertValue -
closingRate_MediumTopRightAlertValue) // 2) + 1))
start = end - (1 * (2**(len(closingRate_FastBottomAlert))))
if (end < 0):
end = 0
if (start < 0):
start = 0
jump = (end - start) // (2**(len(closingRate_FastBottomAlert)))
closingRate_FastBottomAlertValue = transform_fuzzy(
closingRate_FastBottomAlert, jump, start, end)
Distance_CloseTopAlert = current_chromosome[32:37]
Distance_CloseTopAlertValue = transform_fuzzy(
Distance_CloseTopAlert, 50, 0, (50 * (2**len(Distance_CloseTopAlert))))
Distance_FarTopAlert = current_chromosome[37:42]
Distance_FarTopAlertValue = transform_fuzzy(
Distance_CloseTopAlert, 50, (Distance_CloseTopAlertValue + 50),
(Distance_CloseTopAlertValue + 50) +
(50 * (2**len(Distance_CloseTopAlert))))
Distance_CloseBottomAlert = current_chromosome[42:47]
start = (Distance_CloseTopAlertValue + ((
(Distance_FarTopAlertValue - Distance_CloseTopAlertValue) // 2) + 1))
end = Distance_FarTopAlertValue
jump = (end - start) // (2**(len(Distance_CloseBottomAlert)))
Distance_CloseBottomAlertValue = transform_fuzzy(Distance_CloseBottomAlert,
jump, start, end)
Distance_FarBottomAlert = current_chromosome[47:52]
end = (Distance_FarTopAlertValue - ((
(Distance_FarTopAlertValue - Distance_CloseTopAlertValue) // 2) + 1))
start = Distance_CloseTopAlertValue
jump = (end - start) // (2**(len(Distance_FarBottomAlert)))
Distance_FarBottomAlertValue = transform_fuzzy(Distance_FarBottomAlert,
jump, start, end)
#Set variables for Wall feelers, heading and tracking of the agent ##
heading = int(ai.selfHeadingDeg())
tracking = int(ai.selfTrackingDeg())
frontWall = ai.wallFeeler(500, heading)
left45Wall = ai.wallFeeler(500, heading + 45)
right45Wall = ai.wallFeeler(500, heading - 45)
left90Wall = ai.wallFeeler(500, heading + 90)
right90Wall = ai.wallFeeler(500, heading - 90)
left135Wall = ai.wallFeeler(500, heading + 135)
right135Wall = ai.wallFeeler(500, heading - 135)
backWall = ai.wallFeeler(500, heading - 180)
trackWall = ai.wallFeeler(500, tracking)
## Create an array that represents the closing rate of each 45 degree of the full 360 degrees surrounding the agent ##
result_list = []
## Array of the same size, but contains the risk of each direction ##
risk_list = []
for i in range(8):
Degree = tracking + (45 * i)
Speed = ai.selfSpeed()
Distance = ai.wallFeeler(10000, tracking + (45 * i))
result = Closing_Rate(Degree, tracking, Speed, Distance)
result_list.append(result)
### Calculate the Fuzzy membership ###
### 1. Fuzzy Membership For Closing Rate (Speed + Tracking Involved) ###
### 2. Fuzzy Membership For Distance From Walls ###
closing_rate, distance = Closing_Rate(Degree, tracking, Speed,
Distance)
low, medium, fast = Fuzzy_Speed(
closing_rate, closingRate_SlowTopAlertValue,
closingRate_SlowBottomAlertValue,
closingRate_MediumBottomLeftAlertValue,
closingRate_MediumTopLeftAlertValue,
closingRate_MediumTopRightAlertValue,
closingRate_MediumBottomRightAlertValue,
closingRate_FastBottomAlertValue, closingRate_FastTopAlertValue)
close, far = Fuzzy_Distance(distance, Distance_CloseTopAlertValue,
Distance_CloseBottomAlertValue,
Distance_FarBottomAlertValue,
Distance_FarTopAlertValue)
#print("close-far", close, far)
risk = Fuzzy_Risk(low, medium, fast, close, far)
risk_list.append(risk)
## Get the direction in deg that is most risky for the robot as well as the least risky direction ##
max_risk = max(risk_list)
track_risk = (tracking + (risk_list.index(max_risk) * 45) % 360)
min_risk = min(
risk_list
) ## Note: Biase Towards Left Side since min get the first min when risk might be equal ##
####### Getters Variable Regarding Important Information About Enemies ########
##Find the closest ennemy##
enemy = ai.lockClose()
## Get the lockheadingdeg of enemy ##
head = ai.lockHeadingDeg()
## Get the dstance from enemy ##
enemyDist = ai.selfLockDist()
## If the Enemy is Dead ##
if (ai.selfAlive() == 0 and boolean == False):
## Calculate Fitness Current Individual ##
score_previous = score
score_current = ai.selfScore()
fitness_value = fitness(population, count_frame, score_previous,
score_current)
fitness_list.append(fitness_value)
## If it went through the whole population and ready to move to next generation ##
if ((loop + 1) == population_size):
## Output the fitness of population to allow user to see if learning is happening ##
print("Generation:", generation)
print("Agent Fitness:")
print(fitness_list)
print("Average Fitness:", statistics.mean(fitness_list))
print("Best Fitness:", max(fitness_list))
## Finding the optimal chromosome to output it in data file ##
string_maxChromosome = ""
for chrom_max in range(chromosome_size):
string_maxChromosome = string_maxChromosome + str(
population[fitness_list.index(
max(fitness_list))][chrom_max])
## Formatting entire population in a big string to register it in excel file##
string_population = ""
for pop in range(population_size):
for pop_chrom in range(chromosome_size):
string_population = string_population + str(
population[pop][pop_chrom])
if (pop != (population_size - 1)):
string_population = string_population + ","
## Formatting entire population's fitness in a big string to register it in excel file##
string_fitness = ""
for fit in range(len(fitness_list)):
string_fitness = string_fitness + str(fitness_list[fit])
if (fit != (len(fitness_list) - 1)):
string_fitness = string_fitness + ","
## Output Data into Excel File ##
titles = [
"Generation", "Average Fitness", "Best Fitness",
"Population Size", "Chromosome Size", "Crossover Probability",
"Mutation Probability", "Best Chromosome",
"Entire Population Chromosome", "Entire Population Fitness"
]
data = [
generation,
statistics.mean(fitness_list),
max(fitness_list), population_size, chromosome_size,
crossover_prob, mutation_prob, string_maxChromosome,
string_population, string_fitness
]
first_time = Save_Data("Dumpster_Training_Data.xls", 0, titles,
data, first_time)
## Select Population For Next Generation -- Apply Crossover & Mutation ##
new_population = select(population, fitness_list)
new_population = crossover(new_population, chromosome_size,
population_size, crossover_prob)
new_population = mutate(new_population, chromosome_size,
mutation_prob)
population = new_population
loop = 0
count_frame = 0
generation += 1
fitness_list.clear()
### DONE -- QUIT ###
if (generation == generation_size):
quitAI()
## Move to the next individual in population ##
else:
loop += 1
count_frame = 0
boolean = True
else:
## The agent is Alive ##
if (ai.selfAlive() == 1):
## Get the angles on both side between tracking and heading to decide which way to turn ##
dist = (heading - track_risk) % 360
dist2 = (360 - dist) % 360
###### Production System Rules ######
## Turning Rules ##
if (dist <= 130 and dist >= 0 and ai.selfSpeed() > 0
and max_risk >= 75):
ai.turnLeft(1)
elif (dist2 <= 130 and dist2 >= 0 and ai.selfSpeed() > 0
and max_risk >= 75):
ai.turnRight(1)
elif (ai.selfSpeed() <= 10):
ai.thrust(1)
elif (trackWall <= 150):
ai.thrust(1)
##### Bullet Avoidance Commands #####
elif (ai.shotAlert(0) >= 0 and ai.shotAlert(0) <= 50):
if (ai.shotVelDir(0) != -1
and ai.angleDiff(heading, ai.shotVelDir(0)) > 0
and ai.selfSpeed() <= 5):
ai.turnLeft(1)
ai.thrust(1)
elif (ai.shotVelDir(0) != -1
and ai.angleDiff(heading, ai.shotVelDir(0)) < 0
and ai.selfSpeed() <= 5):
ai.turnRight(1)
ai.thrust(1)
elif (ai.shotVelDir(0) != -1
and ai.angleDiff(heading, ai.shotVelDir(0)) > 0
and ai.selfSpeed() > 5):
ai.turnLeft(1)
else:
ai.turnRight(1)
##### Shooting Ennemy Commands #####
elif (enemyDist <= 3000 and heading > (head) and enemyDist != 0
and ai.selfSpeed() > 5):
ai.turnRight(1)
ai.fireShot()
elif (enemyDist <= 3000 and heading < (head) and enemyDist != 0
and ai.selfSpeed() > 5):
ai.turnLeft(1)
ai.fireShot()
## Rules if nothing is happening ##
elif (ai.selfSpeed() < 5):
ai.thrust(1)
else:
ai.thrust(0)
count_frame += 3
boolean = False
def AI_loop():
global frames, frameHistory, generation, current_chrom, population, scores, current_score
print("frame", frames)
#rules
chrom = population[current_chrom][0]
when_to_thrust_speed = convert(chrom[0:4])
front_wall_distance_thrust = convert(chrom[4:8]) * 4
back_wall_distance_thrust = convert(chrom[8:12]) * 2
track_wall_distance_thrust = convert(chrom[12:16]) * 4
left_wall_angle = convert(chrom[16:20]) * 3
right_wall_angle = convert(chrom[20:24]) * 3
back_wall_angle = convert(chrom[24:28]) * 10
left_90_wall_angle = convert(chrom[28:32]) * 6
right_90_wall_angle = convert(chrom[32:36]) * 6
left_back_wall_angle = convert(chrom[36:40]) * 9
right_back_wall_angle = convert(chrom[44:48]) * 9
fuzzy_rate = convert(chrom[48:52])
fuzzy_rate_1 = convert(chrom[52:56])
angle_diff_shoot = convert(chrom[56:60])
shot_alert_distance = convert(chrom[60:64]) * 6
rate = fuzzy_rate
rate1 = fuzzy_rate_1
#Release keys
ai.thrust(0)
ai.turnLeft(0)
ai.turnRight(0)
#Set variables
heading = int(ai.selfHeadingDeg())
tracking = int(ai.selfTrackingDeg())
frontWall = ai.wallFeeler(500, heading)
leftWall = ai.wallFeeler(500, heading + left_wall_angle)
rightWall = ai.wallFeeler(500, heading - right_wall_angle)
left90Wall = ai.wallFeeler(500, heading + left_90_wall_angle)
right90Wall = ai.wallFeeler(500, heading - right_90_wall_angle)
leftbackWall = ai.wallFeeler(500, heading + left_back_wall_angle)
rightbackWall = ai.wallFeeler(500, heading - right_back_wall_angle)
trackWall = ai.wallFeeler(500, tracking)
backWall = ai.wallFeeler(500, heading - back_wall_angle)
#Shooting rules
if ai.aimdir(0) >= 0 and -angle_diff_shoot <= ai.angleDiff(
heading, ai.aimdir(0)) <= angle_diff_shoot and ai.screenEnemyX(
0) >= 0 and ai.screenEnemyY(0) >= 0 and ai.wallBetween(
ai.selfX(), ai.selfY(), ai.screenEnemyX(0),
ai.screenEnemyY(0)) == -1:
ai.fireShot()
#Turn rules
if ai.aimdir(0) >= 0 and ai.angleDiff(
heading, ai.aimdir(0)) > 0 and ai.wallBetween(
ai.selfX(), ai.selfY(), ai.screenEnemyX(0),
ai.screenEnemyY(0)) == -1:
ai.turnLeft(1)
#print("aim turn left")
elif ai.aimdir(0) >= 0 and ai.angleDiff(
heading, ai.aimdir(0)) < 0 and ai.wallBetween(
ai.selfX(), ai.selfY(), ai.screenEnemyX(0),
ai.screenEnemyY(0)) == -1:
ai.turnRight(1)
#print("aim turn right")
elif left90Wall < right90Wall and minmax(10, (rate * ai.selfSpeed()), 100):
ai.turnRight(1)
#print("left90wall")
elif right90Wall < left90Wall and minmax(10, (rate * ai.selfSpeed()), 100):
ai.turnLeft(1)
#print("right90wall")
elif leftbackWall < minmax(5, (rate1 * ai.selfSpeed()), 50):
ai.turnRight(1)
#print("leftbackwall")
elif rightbackWall < minmax(5, (rate1 * ai.selfSpeed()), 50):
ai.turnLeft(1)
#print("rightbackwall")
#base case turn rules
elif leftWall < rightWall:
ai.turnRight(1)
else:
ai.turnLeft(1)
#Thrust rules
if ai.selfSpeed(
) <= when_to_thrust_speed and frontWall >= front_wall_distance_thrust:
ai.thrust(1)
#print("thrust 1")
#dynamic thrust away from walls
elif trackWall < track_wall_distance_thrust:
ai.thrust(1)
#print("thrust 2")
#thrust away from back wall
elif backWall < back_wall_distance_thrust:
ai.thrust(1)
#print("thrust 3")
#Shot avoidance
elif ai.shotAlert(0) >= 0 and ai.shotAlert(
0) <= shot_alert_distance and ai.shotVelDir(
0) != -1 and ai.angleDiff(heading, ai.shotVelDir(0)) > 0:
#print("shotveldir turn left")
ai.turnLeft(1)
ai.thrust(1)
elif ai.shotAlert(0) >= 0 and ai.shotAlert(
0) <= shot_alert_distance and ai.shotVelDir(
0) != -1 and ai.angleDiff(heading, ai.shotVelDir(0)) < 0:
#print("shotveldir turn right")
ai.turnRight(1)
ai.thrust(1)
#tracks frames alive
if ai.selfAlive() == 1:
frames += 1
elif ai.selfAlive() == 0 and frames == 0:
pass
else:
#adds every fitness to a list to avoid the issue of xpilot score never resetting
scores.append(ai.selfScore())
#base case
if len(scores) < 2:
current_score = scores[-1]
else:
current_score = scores[-1] - scores[-2]
if current_score <= 0:
current_score = 2
population[current_chrom][1] = (frames**2) * current_score
print("fitness: ", population[current_chrom][1])
current_chrom += 1
print("current_chrom is: ", current_chrom)
if current_chrom >= len(population):
current_chrom = 0
print("current_chrom is: ", current_chrom)
generation += 1
fitness_scores = fitness(population)
#writes highest fitness agent to file
if generation % 5 == 0:
print("first write")
current_fitness = max(fitness_scores)
best_individual_index = fitness_scores.index(current_fitness)
best_individual = population[best_individual_index][0]
newfile = open("GA_output.txt", "a+")
newfile.write("Generation: ")
newfile.write("\n")
newfile.write(str(generation))
newfile.write("\n")
newfile.write("Best individual: ")
newfile.write("\n")
newfile.write(str(best_individual))
newfile.write("\n")
newfile.write("Best fitness: ")
newfile.write("\n")
newfile.write(str(current_fitness))
newfile.write("\n")
newfile.close()
chrom_pair = top_individuals(population, fitness_scores)
new_pop = []
new_pop += crossover(chrom_pair)
while (len(new_pop) < population_size):
new_pop += crossover(chrom_pair)
population = []
population = new_pop
#writes population to file
if generation % 10 == 0:
print("second write")
popfile = open("GA_pop.txt", "a+")
popfile.write("Generation: ")
popfile.write("\n")
popfile.write(str(generation))
popfile.write("\n")
for i in population:
popfile.write(str(i))
popfile.write("\n")
popfile.write("\n")
popfile.close()
frames = 0
def AI_loop(self):
# print("AI_LOOP")
if ai.selfAlive() == 0:
outputFile = open("fitness.txt", "a")
# outputFile.write(str((self.totalDists/self.counter))+"\t")
outputFile.write(str(int((self.fitness**1.2))) + "\t")
[
print(str("%.5f" % g) + "\t", end="", file=outputFile)
for g in self.chromosome
]
print("\n", end="", file=outputFile)
outputFile.close()
# Release keys
ai.thrust(0)
ai.turnLeft(0)
ai.turnRight(0)
ai.setTurnSpeed(55)
# Heuristics
frontFeelerOffset = 45
perpFeelerOffset = 90
rearFeelerOffset = 135
# turnSpeedMin = 15 # learn range: 4 - 24
turnSpeedMax = 55
speedLimit = 5 # learn range: 2-6
lowSpeedLimit = 2
targetingAccuracy = 4 # 1/2 tolerance in deg for aiming accuracy
shotIsDangerous = 130
# Acquire information
heading = int(ai.selfHeadingDeg())
tracking = int(ai.selfTrackingDeg())
###=== ENEMY FEELERS ===###
# gets angle to enemy
enemyDeg = self.angleToPointDeg(
(ai.selfX(), ai.selfY()), (ai.screenEnemyX(0), ai.screenEnemyY(0)))
enemyWallDistances = []
# maxAngleOffset = 90 # learn range: 30 - 120
# resolution = 5 # learn range: 2 - 10
distAngleTuples = []
# creates tuples of degrees and wallFeelers
for m in (0, self.maxAngleOffset, self.resolution):
distAngleTuples.append(
(enemyDeg - m, ai.wallFeeler(500, int(enemyDeg - m))))
distAngleTuples.append(
(enemyDeg + m, ai.wallFeeler(500, int(enemyDeg + m))))
# gets furthest feeler
maxFeelerAngle = max(distAngleTuples, key=self.returnSecond)
angleToOpenSpace = self.headingDiff(ai.selfHeadingDeg(),
maxFeelerAngle[0])
###=== WALL FEELERS ===###
frontWall = ai.wallFeeler(
self.genericFeelerDist,
heading) # wall feeler for wall directly ahead
leftFrontWall = ai.wallFeeler(
self.genericFeelerDist, heading +
frontFeelerOffset) # wall feeler for wall 45 degrees to the left
rightFrontWall = ai.wallFeeler(
self.genericFeelerDist, heading -
frontFeelerOffset) # wall feeler for wall 45 degrees to the right
leftWall = ai.wallFeeler(
self.genericFeelerDist, heading +
perpFeelerOffset) # wall feeler for wall 90 degrees to the left
rightWall = ai.wallFeeler(
self.genericFeelerDist, heading -
perpFeelerOffset) # wall feeler for wall 90 degrees to the right
backWall = ai.wallFeeler(self.genericFeelerDist, heading -
180) # wall feeler for wall straight back
leftBackWall = ai.wallFeeler(
self.genericFeelerDist, heading +
rearFeelerOffset) # wall feeler for wall 135 degrees to the left
rightBackWall = ai.wallFeeler(
self.genericFeelerDist, heading -
rearFeelerOffset) # wall feeler for wall 135 degrees to the right
trackWall = ai.wallFeeler(
self.genericFeelerDist,
tracking) # wall in front of where ship is moving
# Keep track of all the feeler distances
feelers = [
frontWall, leftFrontWall, rightFrontWall, leftWall, rightWall,
backWall, leftBackWall, rightBackWall, trackWall
]
# Aim assist
leftDir = (heading +
90) % 360 # angle 90 degrees to the left of current heading
rightDir = (heading - 90
) % 360 # angle 90 degrees to the right of current heading
aimer = ai.aimdir(
0
) # direction that the ship needs to turn to in order to face the enemy in degrees
shot = ai.shotAlert(
0
) # returns a danger rating of a shot, the smaller the number the more likely the shot is to hit the ship
enemyX = ai.screenEnemyX(0) # returns the closest enemy's x-coord
enemyY = ai.screenEnemyY(0) # returns the closest enemy's y-coord
selfX = ai.selfX() # returns the ship's x-coord
selfY = ai.selfY() # returns the ship's x-coord
# Fuzzy variable declaration
trackRisk = riskEval(trackWall,
ai.selfSpeed()) #risk of running into trackWall
frontRisk = riskEval(frontWall,
ai.selfSpeed()) #risk of running into frontWall
leftRisk = riskEval(leftWall,
ai.selfSpeed()) #risk of running into leftWall
rightRisk = riskEval(rightWall,
ai.selfSpeed()) #risk of running into rightWall
LFRisk = riskEval(leftFrontWall,
ai.selfSpeed()) #risk of running into leftFrontWall
RFRisk = riskEval(rightFrontWall,
ai.selfSpeed()) #risk of running into rightFrontWall
LBRisk = riskEval(leftBackWall,
ai.selfSpeed()) #risk of running into leftBackWall
RBRisk = riskEval(rightBackWall,
ai.selfSpeed()) #risk of running into rightBackWall
backRisk = riskEval(backWall,
ai.selfSpeed()) #risk of running into backWall
# Compress some wall feelers
sTrack = self.squisher(trackWall)
sLeft = self.squisher(leftFrontWall)
sRight = self.squisher(rightFrontWall)
sLeftStraight = self.squisher(leftWall)
sRightStraight = self.squisher(rightWall)
# output from neural network that tells how much to turn and which direction
turn = self.trainedNeuralNetwork(sTrack, sLeft, sRight, sLeftStraight,
sRightStraight)
###=== THRUST POWER ADJUSTMENT ===#
# Power levels
mfS = self.mfSpeed(ai.selfSpeed())
mfD = self.mfDanger(ai.shotAlert(0))
# if S is high and D is moderate or high:
p1 = max(mfS[2], min(mfD[1], mfD[2]))
# if S is moderate and D is moderate:
p2 = max(mfS[1], mfD[1])
# if S is low and D is high:
p3 = max(mfS[0], mfD[2])
# if S is moderate and D is moderate:
p4 = max(mfS[1], mfD[1])
# if S is low and D is moderate:
p5 = max(mfS[0], mfD[1])
# if S is high and D is low:
p6 = max(mfS[2], mfD[0])
# if S is moderate and D is low:
p7 = max(mfS[1], mfD[0])
# if S is low and D is low:
p8 = max(mfS[0], mfD[0])
consequents = [55, 45, 55, 36, 36, 28, 24, 30]
memberships = [p1, p2, p3, p4, p5, p6, p7, p8]
ai.setPower(self.crispify(memberships, consequents))
if ai.enemyDistance(0) > self.lastDist and ai.enemyDistance(
0) < self.enemyClose:
ai.thrust(1)
elif ai.selfSpeed(
) <= 3 and frontWall >= 200: # if speed is slow and front wall is far away, thrust
ai.thrust(1)
elif trackWall < 60 and frontWall >= 200: # if the track wall is close, thrust
ai.thrust(1)
elif backWall < 20: # if the back wall is close, thrust
ai.thrust(1)
###=== TURNING RULES ===###
# Escape shots
if shot > 0 and shot < 70:
# if a shot is closeby, turn and thrust to avoid
if self.angleDif(rightDir, ai.shotX(0)) < self.angleDif(
leftDir, ai.shotX(0)
) or self.angleDif(rightDir, ai.shotY(0)) < self.angleDif(
leftDir, ai.shotY(0)
): # if shot is coming from the right, turn away and thrust
# print("Turning: avoiding shot")#debug
ai.turnLeft(1)
ai.thrust(1)
elif self.angleDif(leftDir, ai.shotX(0)) < self.angleDif(
rightDir, ai.shotX(0)
) or self.angleDif(leftDir, ai.shotY(0)) < self.angleDif(
rightDir, ai.shotY(0)
): # if shot is coming from the left, turn away and shoot ------> change this shot is just a number
# print("Turning: avoiding shot")#debug
ai.turnRight(1)
ai.thrust(1)
# Turn towards unoccluded enemy
elif aimer >= 0 and self.angleDif(rightDir, aimer) < self.angleDif(
leftDir, aimer) and not self.enemyBehindWall(
0): # if an enemy to the right, turn and shoot it
if ai.screenEnemyX(0) >= 0:
enemyDeg = self.angleToPointDeg(
(ai.selfX(), ai.selfY()),
(ai.screenEnemyX(0), ai.screenEnemyY(0)))
ai.setTurnSpeed(
self.rangeMap(abs(enemyDeg), 0, 180, self.turnSpeedMin,
turnSpeedMax))
else:
enemyDeg = self.angleToPointDeg(
(ai.selfRadarX(), ai.selfRadarY()),
(ai.closestRadarX(), ai.closestRadarY()))
ai.setTurnSpeed(
self.rangeMap(abs(enemyDeg), 0, 180, self.turnSpeedMin,
turnSpeedMax))
# print("Turning: aiming right")#debug
ai.turnRight(1)
elif aimer >= 0 and self.angleDif(leftDir, aimer) < self.angleDif(
rightDir, aimer) and not self.enemyBehindWall(
0): # if an enemy to the left, turn and shoot it
if ai.screenEnemyX(0) >= 0:
enemyDeg = self.angleToPointDeg(
(ai.selfX(), ai.selfY()),
(ai.screenEnemyX(0), ai.screenEnemyY(0)))
ai.setTurnSpeed(
self.rangeMap(abs(enemyDeg), 0, 180, self.turnSpeedMin,
turnSpeedMax))
else:
enemyDeg = self.angleToPointDeg(
(ai.selfRadarX(), ai.selfRadarY()),
(ai.closestRadarX(), ai.closestRadarY()))
ai.setTurnSpeed(
self.rangeMap(abs(enemyDeg), 0, 180, self.turnSpeedMin,
turnSpeedMax))
# print("Turning: aiming left")#debug
ai.turnLeft(1)
#fuzzy avoid walls ahead
elif leftRisk > rightRisk and trackRisk > 0.5: # and min(feelers) < self.nearLimit: #if the left wall and track walls are close, turn right
#if enemyX >=0 and enemyY >= 0 and ai.wallBetween(selfX, selfY, enemyX, enemyY) == -1:
ai.turnRight(1)
# print("Turning: fuzzy right")#debug
elif rightRisk > leftRisk and trackRisk > 0.5: # and min(feelers) < self.nearLimit: #if the right wall and track walls are close, turn left
# if enemyX >=0 and enemyY >= 0 and ai.wallBetween(selfX, selfY, enemyX, enemyY) == -1:
ai.turnLeft(1)
# print("Turning: fuzzy left")#debug
# Turn to open space nearest the angle to the enemy
elif self.enemyBehindWall(0) and min(feelers) > self.nearLimit:
if angleToOpenSpace < 0:
# print("Turning: open space left")#debug
ai.turnLeft(1)
elif angleToOpenSpace > 0:
# print("Turning: open space right")#debug
ai.turnRight(1)
# if neural net value is not between 0.48 and 0.52 then we have to turn right or left
elif not (turn >= 0.43 and turn <= 0.57):
if turn < 0.43: # turn right if value is below 0.43
# print("Turning: neural net right")#debug
ai.turnRight(1)
elif turn > 0.57: # turn left if value is below 0.57
# print("Turning: neural net left")#debug
ai.turnLeft(1)
###=== FIRING RULES ===###
# Restrict firing to reasonably accurate attempts:
# accurate range, enemy not behind wall and enemy close enough
if self.headingDiff(
heading,
ai.aimdir(0)) < targetingAccuracy and not self.enemyBehindWall(
0) and ai.enemyDistance(0) < self.enemyFireDist:
ai.fireShot()
# print("Shot Fired")#debug
# print("Firing Dist: ", self.enemyFireDist)#debug
self.counter += 1
###=== How did we die? and other Fitness Calculations ===###
# Fitness function information
self.totalDists += ai.enemyDistance(0)
if ai.enemyDistance(0) > 0:
self.currentDist = ai.enemyDistance(0)
if self.currentDist < self.lastDist:
self.fitness += 1
self.lastDist = self.currentDist
self.fitness += 1
alive = ai.selfAlive()
message = ai.scanGameMsg(1)
# print(message)#debug
if alive == 0:
self.framesDead += 1
# print(self.framesDead, message)#debug
if self.framesDead == 2:
# print("dead now")#debug
# Ran into wall
if message.find("Beal-Morneault") != -1 and message.find(
"wall") != -1:
print("End of match: wall collision.") #debug
self.fitness -= self.wallPenalty
# Crashed into player
elif message.find("crashed.") != -1:
print("End of match: player collision.") #debug
self.fitness -= self.crashPenalty
# Killed by bullet
elif message.find("Beal-Morneault was") != -1:
print("End of match: killed by opponent.") #debug
self.fitness -= self.killedPenalty
# Killed the opponent
elif message.find("by a shot from Beal-Morneault") != -1:
print("End of match: killed the opponent!") #debug
self.fitness += self.killerBonus
else:
print("End of match: enemy died.")
self.fitness += (ai.selfScore() - ai.enemyScoreId(0)
) * self.scoreDiffBonusFactor
ai.quitAI()
else:
self.framesDead = 0