Path planning design for a wheeled robot: a generative artificial intelligence approach

This article suggests a generative method of path planning design for wheeled robots in narrow streets that uses a high-speed emerging generative AI algorithm—the generative adversarial networks (GANs). The proposed GAN-based architecture efficiently provides accurate footstep planning design for TurtleBot4 on the ROS (Robot Operating System) platform. The designed robot's perception of its surroundings allows it to generate a precise path for navigation during travel. Even though various algorithms, such as A* and RRT* (rapidly exploring random tree), are often employed to plan the path, they need more efficiency in confined spaces. However, deep learning approaches such as GANs have shown remarkable results in solving real-world issues such as image generation and simulation of difficult scenarios in robotics. This article proposes a mechanism to design the GAN algorithm for path generation, thus facilitating the efficient navigation of wheeled robots in complex environments such as narrow streets. According to the experiments, the approach based on GAN works better than traditional algorithms such as heuristic Q-Learning and A*. In comparison to the actual path, it is discovered that the generated path using the path planner based on GAN is approximately 94% accurate.