Hierarchical Manipulation Planning

Abstract

This project presents a hierarchical manipulation planning framework that combines striking, pushing, and pick-and-place motion primitives for efficient manipulation in cluttered environments. The framework selects the most appropriate primitive based on scene context.

Framework

• High-level: Task planner selects optimal primitive sequence
• Mid-level: Motion planner generates collision-free trajectories
• Low-level: Execution with real-time replanning

Results

The hierarchical approach reduces planning time by 60% compared to end-to-end methods while achieving 95% success rate in benchmark manipulation tasks.