MPC-based Coarse-to-Fine Motion Planning for Robotic Object Transportation in Cluttered Environments

TL;DR

This paper introduces a coarse-to-fine motion planning framework using MPC and visual perception for robotic object transportation in cluttered, uncertain environments, enabling real-time, adaptive manipulation.

Contribution

It presents a novel integrated approach combining perception, global planning, and local refinement with a B-spline MPC scheme for cluttered environment manipulation.

Findings

Successfully handles unmodeled, cluttered environments.

Demonstrates robustness and adaptability in experiments.

Supports dynamic replanning and closed-chain kinematics.

Abstract

This letter presents a novel coarse-to-fine motion planning framework for robotic manipulation in cluttered, unmodeled environments. The system integrates a dual-camera perception setup with a B-spline-based model predictive control (MPC) scheme. Initially, the planner generates feasible global trajectories from partial and uncertain observations. As new visual data are incrementally fused, both the environment model and motion planning are progressively refined. A vision-based cost function promotes target-driven exploration, while a refined kernel-perceptron collision detector enables efficient constraint updates for real-time planning. The framework accommodates closed-chain kinematics and supports dynamic replanning. Experiments on a multi-arm platform validate its robustness and adaptability under uncertainties and clutter.