RRT-CS: A free-collision planner for capsule-like SCORBOT by iterated learning

摘要

In this study, we present an enhanced Rapidly-exploring Random Trees (RRT) algorithm integrated with a visual servoing technique for recognizing unknown environments. The robotic platform utilized is the SCORBOT-ER-VII, which consists of five links, servo motors, gearboxes, and an end-effector. Several target objects are used to define the initial position, obstacles, and destination. To evaluate the effectiveness and robustness of our approach, we conducted both numerical simulations and hardware experiments across three test scenarios, ranging from obstacle-free environments to complex obstacle configurations. The results indicate that planning time increases proportionally with scenario complexity. The trajectory smoothing process accounts for less than 10% of the total processing time, while path shortening constitutes one-third, and RRT-based profile generation comprises the remaining two-thirds. These findings clearly demonstrate the efficiency of our approach in terms of computational time, making it well-suited for real-world applications.