Coupled vs. Decoupled Motion Planning for Four-Robotic Arm Manipulation in a Shared Workspace: An Empirical Study
Year: 2025
Type: conference-paper
Abstract: Multi-robot manipulator systems are increasingly used in industrial settings for tasks like coordinated assembly and pick-and-place operations. In such shared workspaces, multiple robotic arms must plan collision-free motions that account for each other’s movements. This multi-robot motion planning problem is computationally challenging, as the joint configuration space grows exponentially with the number of degrees of freedom (DoF). We study motion planning for a four-arm industrial cell (KUKA KR6, KUKA KR10, UR10, UR5) operating in a shared workspace under ROS/Gazebo/MoveIt. We compare a coupled planner in a 24-DoF composite space using OMPL RRT-Connect against a decoupled pipeline that plans per arm and then time-coordinates trajectories with a Detect-and-Pause heuristic. The heuristic discretizes the workspace into voxels, builds a spatio-temporal occupancy matrix, detects inter-arm conflicts in advance, and resolves them by pausing the robot with larger remaining execution time. To ensure a fair comparison, task allocation and grasp configurations are fixed across methods, and identical controller and planning parameters are used. Benchmarks span three clutter levels defined by minimum interobject distance dmin∈{30,20,10}cm. We generate 10 randomized instances per level. Results show the decoupled approach dominates: 100% success at all clutter levels (vs. ∼81.7% for coupled), a −28.9% reduction in median planning time on average, and −26.1% median makespan on average. Gains are largest at 20cm, where interactions are frequent. These findings suggest that time-based coordination on fixed single-arm paths is a simple, robust, and scalable alternative to centralized planning for multi-arm manipulation in shared workspaces.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7128629292
FWCI:
Cites:
Cited by:
Related to: