Obstacle Avoidance Motion Planning and Control of Humanoid Manipulator Based on RRT-Connect and RNN

In complex environments, it remains a challenging issue for robots to simultaneously address obstacle avoidance in both task and joint spaces. To tackle this, we propose an enhancement to the RRT-connect algorithm to optimize path search efficiency, and to generate smooth trajectories by combining greedy pruning with B-spline fitting. Additionally, the obstacle avoidance constraints are reformulated as optimization conditions using quadratic programming, and the inverse kinematics problem is addressed via an RNN solver. Finally, simulation experiments conducted on a self-developed iRobot in the laboratory demonstrate the effectiveness of the enhanced RRT-connect algorithm and the RNN solver based on quadratic programming in achieving obstacle avoidance for point-reaching tasks, offering an effective solution for safe humanoid robot interaction.