Overcoming Dynamic Environments: A Hybrid Approach to Motion Planning for Manipulators

TL;DR

The paper tackles motion planning for robotic manipulators in dynamic, uncertain environments by proposing a hybrid planner that combines a global sampling-based path (via RRT*) in joint space with a local Velocity Potential Field (VPF) for real-time obstacle avoidance. It introduces enhancements to the VPF, including a sigmoid-bounded repulsive field, obstacle-velocity awareness, singularity-avoidance through damped least squares, and mobility-oriented velocity adjustments, all wired into a constrained quadratic program for joint commands. Theoretical guarantees are provided: RRT* path optimality and a completeness/soundness argument for the hybrid planner using Lyapunov analysis, along with a practical switching mechanism between global tracking and local avoidance. Empirical results on the Sawyer manipulator show faster task completion and improved manipulability with safer obstacle avoidance in semi-structured dynamic environments, indicating strong potential for warehousing, manufacturing, and minimally invasive surgical applications.

Abstract

Robotic manipulators operating in dynamic and uncertain environments require efficient motion planning to navigate obstacles while maintaining smooth trajectories. Velocity Potential Field (VPF) planners offer real-time adaptability but struggle with complex constraints and local minima, leading to suboptimal performance in cluttered spaces. Traditional approaches rely on pre-planned trajectories, but frequent recomputation is computationally expensive. This study proposes a hybrid motion planning approach, integrating an improved VPF with a Sampling-Based Motion Planner (SBMP). The SBMP ensures optimal path generation, while VPF provides real-time adaptability to dynamic obstacles. This combination enhances motion planning efficiency, stability, and computational feasibility, addressing key challenges in uncertain environments such as warehousing and surgical robotics.

Figures (13)

• Figure 1: Trajectory executed on a Sawyer robot with a static obstacle
• Figure 2: RRT* search in joint space for 7 DOF manipulators navigating through cluttered environments: the shortest paths to goal with the cost function is the Euclidean distance between joint configurations. Green dots indicate the positions of the end effector along the global paths with a fixed time step, representing the motion's smoothness.
• Figure 3: A single manipulator's link in a scenario with a moving obstacle
• Figure 4: Expected behaviour of the magnitude of the repulsive velocity applied to a link based on the distance to its closest obstacle
• Figure 5: Repulsive velocity in different scenarios: a) Obstacle moves away from the link; b) Obstacle moves along the link; c) Obstacle moves towards the link. Since $\mathbf{d}$ is kept unchanged, the denominator in \ref{['rvf2']} can be simplified to 1 for illustration purposes without the loss of generality.