How can I set some dynamic obstalces in GridWorld?

I am learning to use the reinforcement learning toolbox. I read the tutorial about GridWorld but the positions of the obstacles seem to be fixed and I want to make the obstacles move, for example, one grid at each step. Can the obstalces move?

[Donald Alexander]

Hi,

I have the same problem. I am wondering if I can give a new obstacle value for each individual time step in training? As of now, it seems that the obstacles are fixed for every time step.

I aim to have a moving robot, with a known location at every time step, to be an obstacle.

Thanks

[GU BOCHEN]

Hi Alexander,

I gave up using MATLAB finally, cause I could not solve this problem.

I used Python to build up the environment and here is my codes: GitHub - Adobec/gu-boc-sun: My dissertation codes

Here is my explanation about my codes if you can translate Chinese into your language: 基于强化学习动态避障的Python实现（绝赞摸鱼版）_Adobii的博客-CSDN博客

There are two static and two dynamic obstacles in the environment (gridworld_v1.py), and the Q-learning algorithm is in main.py. Hope this can help.

After that, I found a keyword "puckworld", which is a continuous dynamic environment. I am not sure whether this environment is inside MATLAB RL ToolBox. Maybe you can try to search it.

Best wishes to you

[Donald Alexander]

GU,

Thank you very much for sharing! I appreciate it greatly.

Answer 1

[Priysha LNU]

Hi,

Following are a few answered MATLAB Answers posts with similar queries:

1) https://www.mathworks.com/matlabcentral/answers/437929-making-moving-obstacles-in-an-matlab

2) https://www.mathworks.com/matlabcentral/answers/115540-dynamic-obstacle-in-matlab-environment

The comments and answers provided in these posts might help you resolve this query.

Hope this helps!

Thanks!

DISCLAIMER: These are my own views and in no way depict those of MathWorks.

[GU BOCHEN]

Dear Priysha LNU,

Thanks a lot for your help! It's very helpful!