changelog

机器人足球AI设计比赛

2006-02-28  Xu Yuan  <[e-mail]>
	* the parser (beta version):
	- import/use the SFEXP library
	- use the new parser in WorldModle and Formation
	- the computation is about half of the old parser
	* add some comments in the main.cpp and Parser.h and Parser.cpp
	- "make doc" to generate the doxygen of the project
	* released seu-3d-0.3.c

2006-02-22  Xu Yuan  <[e-mail]>
	* the new log system (beta version):
	- if not need log system, remove all the log code using marco, this is more handiness
	- the more effective log stream
	- auto print the layer, but add LOGBEGIN in the begin of every function

2006-02-12  Xu Yuan  <[e-mail]>
	* the last version before RC2006 qualification
	* the main change are follows:
	- pass, Not only clear ball. But some time not so good.
	- dribble, It is good when dribbling without deceleration, but the beginning is not good.
	- formation, Not change much, still 3-5-2, but more suitable.

2006-01-08  Xu Yuan  <[e-mail]>
	* fixed BUG of calculateOffsideLine
	* add TRAINING_MODE
	- if define this macro, agent and trainer will be compiled into training binary
	- either, the normally run binary
	- it can be easily changed in these two modes by changing one line in Settings.
	- and added PlayerTraining.cpp which contains some agent training functions

2006-01-05  Xu Yuan  <[e-mail]>
	* basic implemented check offside:
	- can check all 22 player is offside
	- but need test with trainer!
	* add some "const" to make code more strong

2006-01-03  Xu Yuan  <[e-mail]>
	* basic implemented jointAction test matlab codes:
	- added
	-- matlab/jointAction/testFsngDribble.m
	-- matlab/jointAction/testDUTturn.m
	-- matlab/jointAction/testFsngCrossPass.m
	-- matlab/share/getVel.m
	-- matlab/share/loadBallData.m

2006-01-02  Xu Yuan  <[e-mail]>
	* import into CVS

2005-12-30  Xu Yuan  <[e-mail]>
	* advance Parser:
	- change in parserNextInt and parserNextFloat
	- using boost/lexical_cast

2005-12-29  Xu Yuan  <[e-mail]>
	* basic implement say action:
	- add SayHear.h and SayHear.cpp
	- but the hear messages from server seems strange,it may be a bug,I have report it.

2005-12-28  Xu Yuan  <[e-mail]>
	* change some compile files:
	- src/Makefile.am converge
	* modify commserver:
	- fix a warnning,add output logs
	

2005-12-13  Xu Yuan  <[e-mail]>
	* implemented commserver:
	- remove the old commserver and lib of spades
	- add ReadBuffFD.h and ReadBuffFD.cpp from spades/shared
	-- this two files and the new commserver of agenttest can run smoothly
	- now the network is also ok :)
	* src/Makefile.am
	- make some changes to suit the codes changing
	- now the trainer is still using oxygen's parser
	* set the version to seu-3d-0.2.a:
	- update ChangeLog, NEWS, INSTALL, TODO

2005-12-12  Xu Yuan  <[e-mail]>
	* implemented parser:
	- the parser can run correctly, 
	- but when parsing some unexpected string, we skip just now
	- it needs more testing!
	
2005-12-11  Xu Yuan  <[e-mail]>
	* remove all codes which bind with lib of rcssserver3d:
	- and just now for the server change, i decided to give up some codes which bind with lib of rcssserver3d
	- after remove all the codes, i need a new parser
	- and use Singleton instead of zeitgeist structure
	- so, we no need the seu-3d-play.rb, and give up using ruby in our agent

2005-12-04  Xu Yuan  <[e-mail]>
	* add Joystick function:
	- now, we can use a joystick to control an agent
	- this function are copyed from RoboLog3D, thanks them!

2005-12-01  Xu Yuan  <[e-mail]>
	* 这些天一直在为AIGame2005作准�
	- 代码已经提交�比赛结果还不知道
	- 这些天的调整已经将能够调整的调整�上层策略采用简单的手工代码
	- 基本能够参加比赛.与以前的代码相比,实力已经大大增强�基本能够战胜清华05全国赛代�
	- 将AIGame2005作为一个milestone,将版本提高到0.2�

2005-11-21  Xu Yuan  <[e-mail]>
	* fix a bug in commserver!
	- it has been changed in agenttest in rcssserver3d.0.3.go

2005-11-21  Xu Yuan  <[e-mail]>
	* basic implemented freedom
	* modify BasicWorldModel's small bug: update myself in team
	* advance defence:
	- now, we use 5 Vs 11, results are follows:
	- BrainStrom05 lost 0~2 goals
	- Aria3d05JP lost 2 goals
	- ZJUBase05CZ lost 1 goal
	- WrightEagle2005CZ lost 1 goal
	- it need more work, special in goalkeeper, and def-position

2005-11-20  Xu Yuan  <[e-mail]>
	* some changes in these days, preparing for AIGame2005:
	- use GA to improve intercept and kickTo
	- implemented clearBall and defendGoal
	- but need more test.

2005-11-10  Xu Yuan  <[e-mail]>
	* modify intercept:
	- but the effect is nothing!
	- it may reach the geometry algorithm's best
	- so, I will not spend more time on it.
	- I will try other algorithm.
	* implemented calculate ground kick force
	- it use a physic formulation instead of loop in SEU-legend
	- it works well.

2005-11-08  Xu Yuan  <[e-mail]>
	* modify class Action and ActionDeque
	- add some new functions
	- they are more easlier to ues
	* implemented whenToKick
	- the base version
	- it return Actions,drive or kick or drive&kick

2005-11-07  Xu Yuan  <[e-mail]>
	* after GA:
	- using GA to adapt var in projectDribbleTo
	- it seems better, but not satistfy me!
	* and do a S function in runTo:
	- makes runTo better
	: today we won the BrainStrom05!(1:0) haha
	: but lose to ZJUBase05 and WrightEagle2005:(
	: some thing need more advance

2005-11-03  Xu Yuan  <[e-mail]>
	* advance the test enviroment in Trainer:
	- random set the ball and the PL
	- reocrd the how much time the PL use when trying intercept ball

2005-11-02  Xu Yuan  <[e-mail]>
	* basic implement the test enviroment:
	- it auto run and log
	- but it is only the beginning, it needs more work.

2005-10-30  Xu Yuan  <[e-mail]>
	* basic implement projectDribbleTo
	- use geometry algorithm:"ZhongChuiXian&QieRuYuan"
	- it begins working, but need more adapt and test.

2005-10-27  Xu Yuan  <[e-mail]>
	* advanced intercept skill:
	- fixed a bug in changeSpeed
	- new a predictInterceptPos
	+- the palyer's drive is considered
	 - so the results will be more accurater
	 - but, it needs more CPU time

2005-10-25  Xu Yuan  <[e-mail]>
	* advanced drive skill:
	- these days, I try many ways to control the agent's drive
	- it is lucky that we make improvmance of drive skill
	- and two functions of drive skills are better than the PID algorithm
	- ( perhaps it is not the wrong of PID, it need cut the whole run to dash and brake time,etc )
	- they are runTo and driveTo
	- howerver, runTo is simpler than driveTo
	- and driveTo can get the time of drive
	- so I will continue to test the time, and use it to improve the intercept.
	+ today, we can win Caspain04 :) most 3 goals!
	- so, set version to seu-3d-0.1.11.c7

2005-10-18  Xu Yuan  <[e-mail]>
	* continue adapt the drive configure:
	- i think the PID paramters can not be changed, it is its best
	- the run position is more accurater
	- but when distacnce is small, and velocity is small
	- this algorithm's drive can not run fast, it takes too long time to accelerate!
	- so, perhaps we need a new algorithm to adapt it.
	* make the predict intercept ball fitter:
	- fix a bug in isFastestOur in WorldModle
	- change the some codes

2005-10-17  Xu Yuan  <[e-mail]>
	* adapt the drive configure:
	- beaucse PID algorithm is uesed in drive;
	- but the paramters is not satistfied!
	- so, i modify these, and it took me more than 5 hours!
	- it is better now, but i am not sure it is the best.

2005-10-13  Xu Yuan  <[e-mail]>
	* use NN in predict ball state:
	- isGroundBall isFallOnGround
	- this two use 6 inputs and i outputs, only 1 layer's neural network
	- this two's results is better than the hand codes.
	- but when trying to use NN in isBallStop, and 18 inputs, 2 layers
	- the results is little better than before, so discard this one.

2005-10-09  Xu Yuan  <[e-mail]>
	* use NN in predict ball state:
	- isGroundBall

2005-10-08  Xu Yuan  <[e-mail]>
	* import nn-utility tool:
	- this is a neural network tool
	- it can create and use NN easily
	- and just now only a test NN was been created, but not well tested.

2005-10-05  Xu Yuan  <[e-mail]>
	* these days test many kinds of method -- to make accurater of ball'position and velocity
	- it is better than before,it is just so.:(
	- the results are: ( data in() when dist<5m)
	- the average position error is about 0.1m(0.05m), and average velocity error is about 5m/s(3m/s)
	- and the max position error is less than 0.5m(0.2m), and max velocity error is about 15m/s
	- set version to seu-3d-0.1.10.c6

2005-10-02  Xu Yuan  <[e-mail]>
	* rewrite the kalmanFilter:
	- move the code kalmanFilter into AgentObject from BasicWorldModel.
	* add some "const"
	
	
2005-10-01  Xu Yuan  <[e-mail]>
	* complete the localization study:
	- the last method is better than the earyer one
	- the average position error is about 0.05m, and average velocity error is about 0.2m/s
	- and the max position error is less than 0.2m, and max velocity error is about 1m/s
	- so, update the version to seu-3d-0.1.9.c6

2005-09-27  Xu Yuan  <[e-mail]>
	* fixed a bug in test m-code:
	- beaucse of getVel's bug when time is the same
	* write a newnewlocaliztion:
	- this newnewlocaliztion is different form the newlocaliztion in order
	- the "newnew" first use the vision message to get a weigthed position
	- and then use the drive power to improve the weigthed position.
	- it seems beeter than the "new", the average position error is about 0.05m.
	- but the velocity is still too error:more than 0.5m/s

2005-09-24  Xu Yuan  <[e-mail]>
	* complete the matlab code for testing localization:
	- make the m-files become one part of this project
	- all uesed m-files contains in ./matlab directory.
	- update the version to seu-3d-0.1.8.c6