def steer_reactive(P, sti, tp, a, t, sx, infleX, infleA, str8ness):
if abs(a) > .6:
return steer_centeralign(P, sti, tp, a)
maxsen = max(t)
ttp = 0
aadj = a
if maxsen > 0 and abs(tp) < .99:
MaxSensorPos = t.index(maxsen)
MaxSensorAng = sangsrad[MaxSensorPos]
sensangF = -.9
aadj = MaxSensorAng * sensangF
if maxsen < 40:
ttp = MaxSensorAng * -P['s2sen'] / maxsen
else:
if str8ness < P['str8thresh'] and abs(infleA) > P['ignoreinfleA']:
ttp = -abs(infleA) / infleA
aadj = 0
else:
ttp = 0
senslimF = .031
ttp = snakeoil.clip(ttp, tp - senslimF, tp + senslimF)
else:
aadj = a
if tp:
ttp = .94 * abs(tp) / tp
else:
ttp = 0
sto = steer_centeralign(P, sti, tp, aadj, ttp)
return speed_appropriate_steer(P, sto, sx)
def throttle_control(P, ai, ts, sx, sl, sy, ang, steer):
ao = ai
if ts < 0:
tooslow = sx - ts
else:
if steer > P['fullstis']:
ts = P['fullstmaxsx']
else:
okmaxspeed4steer = P['stst'] * steer * steer - P['st'] * steer + P[
'stC']
ts = min(ts, okmaxspeed4steer)
tooslow = ts - sx
try:
ao = 2 / (1 + math.exp(-tooslow)) - 1
except:
print "exp overflow:", -tooslow
ao = -1
ao -= abs(sl) * P['slipdec']
spincut = P['spincutint'] - P['spincutslp'] * abs(sy)
spincut = snakeoil.clip(spincut, P['spincutclip'], 1)
ao *= spincut
ww = abs(ang) / P['wwlim']
wwcut = min(ww, .1)
if ts > 0 and sx > 5:
ao -= wwcut
if ao > .8: ao = 1
return ao
def speed_appropriate_steer(P, sto, sx):
if sx > 0:
stmax = max(
P['sxappropriatest1'] / math.sqrt(sx) - P['sxappropriatest2'],
P['safeatanyspeed'])
else:
stmax = 1
return snakeoil.clip(sto, -stmax, stmax)
def traffic_speed_adjustment(os, sx, ts, tsen):
global opHistory
if not opHistory:
opHistory = os
return 0
tsa = 0
mpn = 0
sn = min(os[17], os[18])
if sn > tsen[9] and tsen[9] > 0:
return 0
if sn < 15:
sn = min(sn, os[16], os[19])
if sn < 8:
sn = min(sn, os[15], os[20])
sn -= 5
if sn < 3:
opHistory = os
return -ts
opn = mpn + sn
mpp = mpn - sx / 180
sp = min(opHistory[17], opHistory[18])
if sp < 15:
sp = min(sp, os[16], os[19])
if sp < 8:
sp = min(sn, os[15], os[20])
sp -= 5
opHistory = os
opp = mpp + sp
osx = (opn - opp) * 180
osx = snakeoil.clip(osx, 0, 300)
if osx - sx > 0: return 0
max_tsa = osx - ts
max_worry = 80
full_serious = 20
if sn > max_worry:
seriousness = 0
elif sn < full_serious:
seriousness = 1
else:
seriousness = (max_worry - sn) / (max_worry - full_serious)
tsa = max_tsa * seriousness
tsa = snakeoil.clip(tsa, -ts, 0)
return tsa
def drive(c, tick):
S, R, P = c.S.d, c.R.d, c.P
print(P)
global target_speed
global lap
global prev_distance_from_start
global learn_final
global badness
badness = S['damage'] - badness
skid = skid_severity(P, S['wheelSpinVel'], S['speedX'])
if skid > 1:
badness += 15
if car_might_be_stuck(S['speedX'], S['angle'], S['trackPos']):
S['stucktimer'] = (S['stucktimer'] % 400) + 1
if car_is_stuck(S['speedX'], S['stucktimer'], S['angle'],
S['trackPos'], S['track'][9], target_speed):
badness += 100
R['brake'] = 0
if target_speed > 0:
target_speed = -40
else:
target_speed = 40
else:
S['stucktimer'] = 0
if S['z'] > 4:
badness += 20
infleX, infleA, straightness = track_sensor_analysis(
S['track'], S['angle'])
if target_speed > 0:
if c.stage:
if not S['stucktimer']:
target_speed = speed_planning(P, S['distFromStart'],
S['track'], S['trackPos'],
S['speedX'], S['speedY'],
R['steer'], S['angle'], infleX,
infleA)
target_speed += jump_speed_adjustment(S['z'])
if c.stage > 1:
target_speed += traffic_speed_adjustment(
S['opponents'], S['speedX'], target_speed, S['track'])
target_speed *= damage_speed_adjustment(S['damage'])
else:
if lap > 1 and T.usable_model:
target_speed = speed_planning(P, S['distFromStart'],
S['track'], S['trackPos'],
S['speedX'], S['speedY'],
R['steer'], S['angle'], infleX,
infleA)
target_speed *= damage_speed_adjustment(S['damage'])
else:
target_speed = 50
target_speed = min(target_speed, 333)
caution = 1
if T.usable_model:
snow = T.section_in_now(S['distFromStart'])
snext = T.section_ahead(S['distFromStart'])
if snow:
if snow.badness > 100: caution = .80
if snow.badness > 1000: caution = .65
if snow.badness > 10000: caution = .4
if snext:
if snow.end - S['distFromStart'] < 200:
if snext.badness > 100: caution = .90
if snext.badness > 1000: caution = .75
if snext.badness > 10000: caution = .5
target_speed *= caution
if T.usable_model or c.stage > 1:
if abs(S['trackPos']) > 1:
s = steer_centeralign(P, R['steer'], S['trackPos'], S['angle'])
badness += 1
else:
s = steer_reactive(P, R['steer'], S['trackPos'], S['angle'],
S['track'], S['speedX'], infleX, infleA,
straightness)
else:
s = steer_centeralign(P, R['steer'], S['trackPos'], S['angle'])
R['steer'] = s
if S['stucktimer'] and S['distRaced'] > 20:
if target_speed < 0:
R['steer'] = -S['angle']
if c.stage > 1:
if target_speed < 0:
target_speed *= snakeoil.clip(S['opponents'][0] / 20, .1, 1)
target_speed *= snakeoil.clip(S['opponents'][35] / 20, .1, 1)
else:
R['steer'] = speed_appropriate_steer(
P, traffic_navigation(S['opponents'], R['steer']),
S['speedX'] + 50)
if not S['stucktimer']:
target_speed = abs(target_speed)
slip = find_slip(S['wheelSpinVel'])
R['accel'] = throttle_control(P, R['accel'], target_speed, S['speedX'],
slip, S['speedY'], S['angle'], R['steer'])
if R['accel'] < .01:
R['brake'] = brake_control(P, R['brake'], S['speedX'], S['speedY'],
target_speed, skid)
else:
R['brake'] = 0
R['gear'], R['clutch'] = automatic_transimission(P, S['rpm'], S['gear'],
R['clutch'], S['rpm'],
S['speedX'], target_speed,
tick)
R['clutch'] = clutch_control(P, R['clutch'], slip, S['speedX'],
S['speedY'], S['gear'])
if S['distRaced'] < S['distFromStart']:
lap = 0
if prev_distance_from_start > S['distFromStart'] and abs(S['angle']) < .1:
lap += 1
prev_distance_from_start = S['distFromStart']
if not lap:
T.laplength = max(S['distFromStart'], T.laplength)
elif lap == 1 and not T.usable_model:
learn_track(R['steer'], S['angle'], S['track'], S['distFromStart'])
elif c.stage == 3:
pass
else:
if not learn_final:
learn_track_final(T.laplength)
T.post_process_track()
learn_final = True
if T.laplength:
properlap = S['distRaced'] / T.laplength
else:
properlap = 0
if c.stage == 0 and lap < 4:
T.record_badness(badness, S['distFromStart'])
S['targetSpeed'] = target_speed
target_speed = 70
badness = S['damage']
return
def jump_speed_adjustment(z):
offtheground = snakeoil.clip(z - .350, 0, 1000)
jsa = offtheground * -800
return jsa
def drive(c, genoma=None):
S = c.S.d
R = c.R.d
# S: {'angle': 0.0467207, 'curLapTime': 2.468, 'damage': 0.0, 'distFromStart': 2047.54, 'distRaced': 14.9774,
# 'fuel': 94.0, 'gear': 1.0, 'lastLapTime': 0.0,
# 'opponents': [200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0,
# 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0,
# 4.85916, 200.0, 200.0, 200.0, 200.0, 200.0, 200.0, 10.4405], 'racePos': 1.0, 'rpm': 5792.73,
# 'speedX': 48.7787, 'speedY': 0.743483, 'speedZ': 0.0274631,
# 'track': [5.37889, 5.66359, 6.49984, 8.44336, 14.2813, 29.5085, 65.08, 200.0, 100.793, 54.695, 37.7275, 28.963,
# 23.649, 20.1121, 15.7716, 12.4104, 10.6624, 9.82549, 9.63669], 'trackPos': 0.283664,
# 'wheelSpinVel': [41.4835, 40.4579, 43.3378, 42.535], 'z': 0.344934, 'focus': [-1.0, -1.0, -1.0, -1.0, -1.0]}
# R: {'accel': 0.5600000000000006, 'brake': 0, 'clutch': 0, 'gear': 1, 'steer': 0.12010916050557581,
# 'focus': [-90, -45, 0, 45, 90], 'meta': 0}
# # Damage Control
# target_speed-= S['damage'] * .05
# if target_speed < 25: target_speed= 25
#
# # Steer To Corner
# R['steer']= S['angle']*10 / PI
# # Steer To Center
# R['steer']-= S['trackPos']*.10
# R['steer']= clip(R['steer'],-1,1)
# # Throttle Control
# if S['speedX'] < target_speed - (R['steer']*50):
# R['accel']+= .01
# else:
# R['accel']-= .01
# if S['speedX']<10:
# R['accel']+= 1/(S['speedX']+.1)
target_speed = 100
# Steer To Corner
R['steer'] = S['angle'] * 10 / PI
# Steer To Center
R['steer'] -= S['trackPos'] * .10
# Throttle Control
if S['speedX'] < target_speed - (R['steer'] * 50):
R['accel'] += .01
else:
R['accel'] -= .01
if S['speedX'] < 10:
R['accel'] += 1 / (S['speedX'] + .1)
winner_net = neat.nn.FeedForwardNetwork.create(genoma, config)
improved_frontal = max(S['track'][8], S['track'][9], S['track'][10])
inputs = [
S['angle'], S['speedX'], S['speedY'], S['speedZ'], S['trackPos'],
S['track'][0], S['track'][3], S['track'][6], improved_frontal,
S['track'][12], S['track'][15], S['track'][18]
]
delta_accel, delta_brake, delta_steer = winner_net.activate(inputs)
R['accel'] += delta_accel
R['brake'] += delta_brake
R['steer'] += delta_steer
# Traction Control System
if ((S['wheelSpinVel'][2] + S['wheelSpinVel'][3]) -
(S['wheelSpinVel'][0] + S['wheelSpinVel'][1]) > 5):
R['accel'] -= .2
R['accel'] = clip(R['accel'], 0, 1)
# Automatic Transmission
R['gear'] = 1
if S['speedX'] > 50:
R['gear'] = 2
if S['speedX'] > 80:
R['gear'] = 3
if S['speedX'] > 110:
R['gear'] = 4
if S['speedX'] > 140:
R['gear'] = 5
if S['speedX'] > 170:
R['gear'] = 6
#print("R: ", R)
return
