search

ethz-asl / rl-navigation

BSD 3-Clause "New" or "Revised" License
63 stars 28 forks source link

pre-training for the Imitation learning #5

Closed lukewenMX closed 4 years ago

lukewenMX commented 4 years ago

Hi, all. I am a Ph.D. student from NTU, Singapore and also work in the mapless navigation research direction. I must admit this is an amazing project, however, I am struggling with a problem. Are you training both Actor and Critic neural networks in the Imitation learning process? Look forward to your soon response. Thanks.

Weeeesen commented 4 years ago

@lukewenMX hi, I am an undergraduate in NUS, doing a project based on this paper as well and having some troubles including loading the pre-trained weights. If you don't mind, are there any way I could reach out to you for further discussions?

pfmark commented 4 years ago

The pre-training weights are only used to initialize the actor. I admit, this could be stated more clearly in the paper. Initializing the critic with the same weights would not make sense. The critic is trained from scratch in every training run.

minded-hua commented 4 years ago

@lukewenMX @kermeed I ran the code but met a problem. The problem is that the loaded robot model can't run within the map. I guess it's caused by that the coordinate system's origin of /odom and /map is mismatching, however I can't solve it. rviz

Did you met the problem? Looking forward to your reply. Thanks.

sumingzs11 commented 4 years ago

@minded-hua Hello, I want to ask about the operation of the experiment. Whether I initialize or retrain, I never succeed once, and the robot always fails to reach the target point. I think it's probably a problem with my 2D simulator, stageros. I used the stageros simulator that came with ROS kinetic, but its clock was so slow that the training time was too long. So, I modified the time parameters, but I also found a problem, that is, the simulator can't send a stall signal, I don't know if it's because of this that my robot can't reach the target.

lukewenMX commented 4 years ago

@lukewenMX hi, I am an undergraduate in NUS, doing a project based on this paper as well and having some troubles including loading the pre-trained weights. If you don't mind, are there any way I could reach out to you for further discussions?

minded-hua commented 4 years ago

@sumingzs11 Do you have a WeChat ID? We can talk with WeChat.

sumingzs11 commented 4 years ago

Sorry, I was just having lunch. If you are willing to communicate on WeChat, that is really great. My wechat id is  ibelieveyouonly

---Original--- From: "YoungerSky"<notifications@github.com> Date: Wed, Sep 2, 2020 12:12 PM To: "ethz-asl/rl-navigation"<rl-navigation@noreply.github.com>; Cc: "Mention"<mention@noreply.github.com>;"sumingzs11"<847366070@qq.com>; Subject: Re: [ethz-asl/rl-navigation] pre-training for the Imitation learning (#5)

@sumingzs11 Do you have a WeChat ID? We can talk with WeChat.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

sumingzs11 commented 4 years ago

------------------ 原始邮件 ------------------ 发件人: "ethz-asl/rl-navigation" <notifications@github.com>; 发送时间: 2020年9月2日(星期三) 中午12:12 收件人: "ethz-asl/rl-navigation"<rl-navigation@noreply.github.com>; 抄送: "苏铭"<847366070@qq.com>;"Mention"<mention@noreply.github.com>; 主题: Re: [ethz-asl/rl-navigation] pre-training for the Imitation learning (#5)

@sumingzs11 Do you have a WeChat ID? We can talk with WeChat.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.