readme

这是bs球队的仿真2D比赛程序

WARNING:
found a bug, which I can't easily remove.
If you choose appType: 2, but load a latticemap (appInitFN) and probably
also if you load something completely different. The player will crush without
any message. (n++ doesn't give a message). So make sure that appInitFN really is
the filename of a fitting network.

The parameters for the two learned approximators:
lmTd0Gree10xx60000.lmp  
######################################################################
# This file was created by the following run of policy_one_vs_goalie2
# use_handicap_goalshot_test 1
# handicap: 2.000000
# safety_handicap_sub: 1.000000
# risky_dis_to_pole_5: 0.800000
# risky_dis_to_pole_30: 1.900000
# reconsider_goalshot: 1
# wait_and_see_patience: 100
# wait_and_see_clearanceline: 15.000000
# kickrush_param: 1.000000
# flank_param: 0.100000
# use_handicap_selfpasses: 1
# selfpasses_handicap: 2.000000
# selfpasses_safety_level: 3.000000
# selfpasses_SHORT_safety_level: 3.000000
# selfpasses_SHORT_max_speed: 0.710000
# use_goalshot: 1
# appType: 0
# reprType: 0
# appInitFN: /home/deisen/fappTest/initLMPovg8angKlein.lmp
# saveFNTrunc: /home/deisen/robocup/learn1gg1/uni/vFeldMelbourneGoalie8dim/LM/td0Gree10akt140/xx
# learnRate: 0.300000
# randomEvery: 10
# trainEveryTraj: -1
# trajAmount: 8000
# updateMode: 0
# nrOfIterations: 1
# saveEvery: 30000
# printPatterns: 0
# printLearn: 0
# evalEvery: 500
# evalAmount: 50
# kickActionParams: {dir: (-3.1, 2.8, 20), strength: (0.4, 1.0, 7)}
# Boltz-T: 0.100000
######################################################################

rpropTd1Gree10Iter30xx1000.net 
######################################################################
# This file was created by the following run of policy_one_vs_goalie2
# use_handicap_goalshot_test 1
# handicap: 2.000000
# safety_handicap_sub: 1.000000
# risky_dis_to_pole_5: 0.800000
# risky_dis_to_pole_30: 1.900000
# reconsider_goalshot: 1
# wait_and_see_patience: 100
# wait_and_see_clearanceline: 15.000000
# kickrush_param: 1.000000
# flank_param: 0.100000
# use_handicap_selfpasses: 1
# selfpasses_handicap: 2.000000
# selfpasses_safety_level: 3.000000
# selfpasses_SHORT_safety_level: 3.000000
# selfpasses_SHORT_max_speed: 0.710000
# use_goalshot: 1
# appType: 2
# reprType: 0
# appInitFN: /home/deisen/fappTest/npp_initNet8_20_1Scaled.net
# saveFNTrunc: /home/deisen/robocup/learn1gg1/uni/vFeldMelbourneGoalie8dim/RPROP/td1Gree10Iter30/xx
# learnRate: 0.200000
# randomEvery: 10
# trainEveryTraj: 300
# trajAmount: 10000
# updateMode: 1
# nrOfIterations: 30
# saveEvery: 20
# printPatterns: 0
# printLearn: 1
# evalEvery: 500
# evalAmount: 50
# kickActionParams: {dir: (-3.1, 2.8, 20), strength: (0.4, 1.0, 7)}
# Boltz-T: 0.100000
######################################################################