In this study, we provide a visualization tool that can teach itself how to drive an autonomous vehicle. We used a limited number of awards in conjunction with vehicle control policies. By utilizing our simulated trajectories across an environment, we are going to provide fresh training data that enables different new local trajectories to be followed by virtual agents, each with a different perspective of the scene and stable with the road appearance and its complexities. Without using any human control labels during training, we'll demonstrate how policies picked up in our virtual world can be applied to and navigate across previously invisible pathways. Our findings are in line with the learned procedure applied to an entirely autonomous vehicle, such as in conditions that had never been encountered before, such as new highways and unusual, intricate, close calls. Our techniques are extendable, utilizing reinforcement learning, and applying it to conditions needing a powerful physical operation and effective perception