Dear Author,I am currently researching how to incorporate lane selection into an intersection environment. I aim to create a multi-lane environment where ego vehicle, upon reaching the stop line of an intersection, can choose a lane using reinforcement learning or numerical optimization methods to better reflect real-world vehicle behavior. I have noticed that the original environment appears to have predetermined vehicle paths. Do I need to modify the behavior of ego vehicle at a more fundamental level in the code? I would like to know if such modifications are feasible and seek your assistance. Thank you!