A Collision-Aware Motion Planning Framework With Goal State Generation for Aerial Manipulation in Constrained Environments
Zhaopeng Zhang, Hai Yu, Wei He, Yongchun Fang, Jianda Han, and Xiao Liang
IEEE Transactions on Industrial Electronics, early access, 2026
This work proposes a collision-aware whole-body motion planning framework for aerial manipulation in constrained environments. Given a target end-effector pose, the framework selects a favorable goal state, searches for a whole-body path, and optimizes a dynamically consistent trajectory. A link-level collision model accounts for the geometry of both the multi-rotor and the robotic arm, enabling manipulation in confined spaces. The framework is validated in both simulation and real-world experiments.
Key Features
- Collision-Aware Goal State Generation: Exploits kinematic redundancy to select a goal state that minimizes a collision-avoidance margin while satisfying the target end-effector position and orientation
- Whole-Body Path Searching: Uses a goal-state-biased bidirectional RRT* search with a weighted configuration-space distance to connect the initial state to the generated goal state
- Link-Level Collision Modeling: Represents the multi-rotor and individual arm links with multiple sets of spheres, with an adjustment to account for platform tilt during motion
- Dynamically Consistent Trajectory Optimization: Uses the complete system’s flat outputs, including its center of mass, to account for dynamic coupling while incorporating collision avoidance and motion feasibility constraints