开源项目 TORCS经验

安装部分

sudo apt-get install xautomation报错

The following packages have unmet dependencies:
 electron-ssr : Depends: libappindicator1 but it is not going to be installed
E: Unmet dependencies. Try 'apt-get -f install' with no packages (or specify a solution).

按照提示执行 sudo apt-get -f install 报错can’t fetch 清华源
开始换源：
1、备份原始文件

sudo cp /etc/apt/sources.list /etc/apt/sources.list.backup

2、打开文件

vim /etc/apt/sources.list

3、阿里源也不行，换成东北大学源

deb-src http://mirror.neu.edu.cn/ubuntu/ xenial main restricted #Added by software-properties
deb http://mirror.neu.edu.cn/ubuntu/ xenial main restricted
deb-src http://mirror.neu.edu.cn/ubuntu/ xenial restricted multiverse universe #Added by software-properties
deb http://mirror.neu.edu.cn/ubuntu/ xenial-updates main restricted
deb-src http://mirror.neu.edu.cn/ubuntu/ xenial-updates main restricted multiverse universe #Added by software-properties
deb http://mirror.neu.edu.cn/ubuntu/ xenial universe
deb http://mirror.neu.edu.cn/ubuntu/ xenial-updates universe
deb http://mirror.neu.edu.cn/ubuntu/ xenial multiverse
deb http://mirror.neu.edu.cn/ubuntu/ xenial-updates multiverse
deb http://mirror.neu.edu.cn/ubuntu/ xenial-backports main restricted universe multiverse
deb-src http://mirror.neu.edu.cn/ubuntu/ xenial-backports main restricted universe multiverse #Added by software-properties
deb http://archive.canonical.com/ubuntu xenial partner
deb-src http://archive.canonical.com/ubuntu xenial partner
deb http://mirror.neu.edu.cn/ubuntu/ xenial-security main restricted
deb-src http://mirror.neu.edu.cn/ubuntu/ xenial-security main restricted multiverse universe #Added by software-properties
deb http://mirror.neu.edu.cn/ubuntu/ xenial-security universe
deb http://mirror.neu.edu.cn/ubuntu/ xenial-security multiverse

全覆盖就行
4、重新执行apt-get -f install

ubuntu E: Package ‘libpng12-dev’ has no installation candidate

18.04后’libpng12-dev’被废弃，改成libpng-dev。

运行torcs

1. 出现报错
Compatibility mode, properties unknown.
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 151 (GLX)
Minor opcode of failed request: 24 (X_GLXCreateNewContext)
Value in failed request: 0x0
Serial number of failed request: 31
Current serial number in output stream: 32

google一下发现似乎是显卡驱动有问题，在尝试更新驱动无效后，发现是服务器的VNC的bug，在MobaXterm上直接运行torcs即可。（Gobigger也有同样的显示bug。）
2. 帧数一直是个位数，在查看源代码良久无效，遂换到20.04上重新安装，正常了。百思不得其解，望如有解决方法请告知，不胜感激。

开始训练

1. 在成功运行torcs后开始ddpg的训练，运行ddpg报错Shapes (3, 2, 1) and () are incompatible，原因是安装的TensorFlow版本较高，把虚拟环境中keras/backend/tensorflow_backend.py文件修改第716行，改成tf.concat的两个参数换一下顺序即可。

2. 报错：ImportError: cannot import name collect_trainable_weights
原因：keras版本过高，重新pip install keras==1.1.0
或者

• 注释掉ddpg.py中的from keras.engine.training import collect_trainable_weights
  修改ActorNetwork.py，将from keras.initializations import normal, identity中的initializtion改为initializers,注释掉from keras.engine.training import collect_trainable_weights
  修改CriticNetworks.py，keras.initializtion改为keras.initializers, 注释掉from keras.engine.training

3. 1）跑起来总有一种歪歪扭扭的感觉；改进构想：调整参数或调整奖励值。
2）车子经常撞到赛道边缘然后就卡住不动了；改进构想：新增倒车动作或当当前状态与前一状态的位移差小于阈值，运行reset。

参考：“Simulated Car Racing Championship Competition Software Manual”