Combined workspace monitoring and collision avoidance for mobile manipulators

Zube, Angelika · 2015 · Crossref

DOI: 10.1109/etfa.2015.7301526

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper presents a Nonlinear Model Predictive Control (NMPC) approach for fixed-base and mobile manipulators that integrates Cartesian task execution with real-time collision avoidance. The research addresses the challenge of enabling safe human-robot interaction in shared, cluttered workspaces, where robots must avoid collisions with static and dynamic obstacles—including self-collisions—without interrupting their primary tasks. Unlike reactive strategies that stop robot motion upon detecting proximity, this method utilizes robot redundancy to perform evasive motions while continuing to follow reference trajectories. The methodology combines a 3D workspace monitoring system with an NMPC controller. The monitoring system fuses data from multiple depth sensors (laser scanners and Kinect sensors) to create a conservative 3D octree representation of obstacles, accounting for occlusions behind the robot and other objects. The NMPC controller computes optimal joint velocities to track the desired end-effector pose while satisfying inequality constraints for joint limits, self-collision avoidance, and obstacle clearance. To ensure real-time performance, the algorithm restricts collision avoidance calculations to the most relevant obstacles—those closest to specific robot parts—and utilizes a reachability grid to filter out obstacles the robot cannot physically reach within the prediction horizon. The approach was validated through simulations and experiments on a 10-degree-of-freedom omnidirectional mobile manipulator (KUKA OmniRob with LWR IV). Simulation results compared different strategies for selecting relevant obstacles, demonstrating that considering the closest obstacle for each individual robot sphere yielded the best balance of safety and computational efficiency. In experimental trials involving a human walking into the robot’s path, the system successfully executed evasive maneuvers, such as moving the mobile base backward, to maintain safety. The robot maintained a minimum distance of 27 cm from the human, with a maximum end-effector position deviation of 3.2 cm and orientation deviation of 3.1 degrees from the reference trajectory. The control algorithm operated with a computation time of under 60 ms per step, meeting the real-time requirements for a 100 ms sampling interval. The significance of this work lies in its ability to provide foresightful, continuous collision avoidance for mobile manipulators in dynamic environments. By integrating comprehensive 3D workspace monitoring with predictive control, the method ensures safety without the task interruptions associated with traditional stop-and-go strategies. The results demonstrate that efficient collision avoidance is achievable in real-time by intelligently selecting relevant obstacles, making the approach suitable for complex industrial applications involving shared human-robot workspaces.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• time to collision looming