How to add custom environment for Reinforcement learning toolbox?

I want to make a 3d environment representing a neighborhood containing some blocks as the buildings. I want to use this model as the environment in the toolbox for the agent to interact with and find the shortest path. How should I do that? I'm having difficulties defining this using classes.

Answer 1

[Emmanouil Tzorakoleftherakis]

Why do you need a 3d world for this problem? Unless you consider the z dimension (e.g. if you do planning for UAVs), you only need a 2d env. Even if you were to do it for visualization, I wouldn't recommend training in the 3d world since it would only make training slower. I would start with a grid world or some occupancy grid that you can tailor to match the 3d world you have in mind:

https://www.mathworks.com/help/reinforcement-learning/ug/train-q-learning-agent-to-solve-basic-grid-world.html

https://www.mathworks.com/help/robotics/ug/avoid-obstacles-using-reinforcement-learning-for-mobile-robots.html

[shahin sarhan]

That's the problem, my agent is a UAV. I want to find the optimum path by minimising the noise emiitted from the drone. Thats why I need an interactive 3D space.

[Emmanouil Tzorakoleftherakis]

You can also create 3d occupancy maps like this:

https://www.mathworks.com/help/uav/ug/generate-random-3-d-occupancy-map-for-uav-motion-planning.html

https://www.mathworks.com/help/nav/ref/occupancymap3d.html

and turn them into RL training environments by following the example I shared earlier:

https://www.mathworks.com/help/robotics/ug/avoid-obstacles-using-reinforcement-learning-for-mobile-robots.html