Predictive Multi-Agent Coordination for Adaptive and Human-Aware Collaborative Assembly Robotics
Abstract
Planning autonomous mobile robots in real-time and unpredictable and dynamic environments is a significant challenge. A*, D-Lite, RRT, and Velocity Obstacle (VO) initially based approaches are traditional in that either they utilize a fixed world map or are reactive-only. Global planners work well in nominal settings, but cannot adjust under dynamic settings, whereas reactive local approaches such as Dynamic Window Approach (DWA) and Artificial Potential Fields (APF) are quicker but tend to give wiggly or wavy paths. More computationally expensive systems such as Model Predictive Control (MPC) also place extra processing load on the on-board computer to the point of constraining speed and safety. In order to overcome these shortcomings, a hierarchical hybrid path planning algorithm that integrates Dynamic Wavefront Decomposition (DWD) with a Predictive Local Replanning Module (PLRM) is suggested. To achieve the continuity of global trajectories, DWD dynamically divides the environment and calculates parallel costmaps, and predicts the motion of obstacles to produce smoother and adaptive local paths according to PLRM. Experimental findings indicate that the DWD-PLRM framework can minimize the replanning time by 65.9 percent relative to D*-Lite and 66 percent relative to A+ and ensures that the latencies are kept below 100 ms. It has a 92% obstacle avoidance success rate which is 10-percent better, as compared to current techniques. The presented solution provides an adapted and predictive, computationally efficient, and real-time autonomous navigation solution, a unification of the global optimality of structured planners with the flexibility of reactive systems to ensure reliable multi-robot behaviour.