The robotics project would start like this: in Pybullet, build a simulated ground vehicle that is supposed to go from an initial location to some goal location, and there are random static obstacles in between. Then have this environment connected with OpenAI Gym so that you can run standard policy gradient RL algorithms such as PPO and SAC to train the controller for the car so that it goes from the initial state to the goal and avoid the obstacles.
Getting familiar with these tools may require quite some effort, so take your time and let me know when you have something working.
(something like this https://urldefense.com/v3/__https://www.youtube.com/watch?v=tozzb6xC1LI__;!!Mih3wA!Buejbwp2bgHIMcIvfX7xWLQr6D-qXoKnlnkY_rIRrQEL0KTlHL8E_qtZHCSQTc3S3kCAQtqNhRyLrw$ )