Hello everyone,
I am trying to achieve obstacle avoidance using raycasts in a more compact way but it doesnt seem to work and I cant figure out why. In my pipeline, I get ‘Obstacles Distances’ which is an array of floats containing 50 floats. These are the hit distances. So there are indices from 0-49 with values ranging from 0 to 1 (Normalized). I add penalty when the agent gets close to obstacles and i can see that it gets the penalties, however the agent never tries to avoid obstacles and just keeps following the spline.
The only thing i can think about is how i am giving the observation to the agent as can be seen in the images below. Is this approach ok? or am i missing something?
