Applying reinforcement learning techniques in real robotics simulation

This is based on the works of https://arxiv.org/pdf/2002.04349.pdf this paper but instead of using LIDAR, I tried using an array of ultrasonic sensors

A dueling deep neural network is used and is based on the PyTorch framework the results are highly arbitary, still under hyperparameter tuning. The reward system is based on the paper though they havn't fully explained the academic approach in detail hence I have applied my own academic style approach.

I have also included models used in the simulation, you can donwload and change the address in config file to work with them.