COMPASS: Confined-space Manipulation Planning with Active Sensing Strategy
Qixuan Li, Chen Le, Dongyue Huang, Jincheng Yu, Xinlei Chen
AI summary
Problem
Manipulation in confined spaces fails due to severe perception occlusion and tight kinematic constraints, yet existing methods assume full observability or known targets and lack integrated exploration-manipulation strategies.
Approach
The framework uses a wrist-camera scan for immediate safety, a manipulation-aware utility planner to select viewpoints that balance information gain and grasp feasibility, and constrained optimization to generate collision-free grasp poses.
Key results
- Proposed COMPASS multi-stage framework for confined-space exploration and manipulation
- Developed MUE-RRT planner balancing information gain, manipulability, and motion cost
- Introduced a progressively challenging simulation benchmark for confined-space tasks
- Achieved 24.25% higher manipulation success rate in simulations and validated real-world performance
Why it matters
Enables autonomous robots to safely navigate, perceive, and manipulate objects in tight, cluttered environments where traditional methods fail.
Abstract
Manipulation in confined and cluttered environ- ments remains a significant challenge due to partial observabil- ity and complex configuration spaces. Effective manipulation in such environments requires an intelligent exploration strategy to safely understand the scene and search the target. In this paper, we propose COMPASS, a multi-stage exploration and manipulation framework featuring a manipulation-aware sampling-based planner. First, we reduce collision risks with a near-field awareness scan to build a local collision map. Additionally, we employ a multi-objective utility function to find viewpoints that are both informative and conducive to subsequent manipulation. Moreover, we perform a constrained manipulation optimization strategy to generate manipulation poses that respect obstacle constraints. To systematically evalu- ate method’s performance under these difficulties, we propose a benchmark of confined-space exploration and manipulation containing four level challenging scenarios. Compared to explo- ration methods designed for other robots and only considering information gain, our framework increases manipulation suc- cess rate by 24.25% in simulations. Real-world experiments demonstrate our method’s capability for active sensing and manipulation in confined environments.