A Collision-Free MPC for Whole-Body Dynamic Locomotion and Manipulation (2202.12385v1)

Abstract: In this paper, we present a real-time whole-body planner for collision-free legged mobile manipulation. We enforce both self-collision and environment-collision avoidance as soft constraints within a Model Predictive Control (MPC) scheme that solves a multi-contact optimal control problem. By penalizing the signed distances among a set of representative primitive collision bodies, the robot is able to safely execute a variety of dynamic maneuvers while preventing any self-collisions. Moreover, collision-free navigation and manipulation in both static and dynamic environments are made viable through efficient queries of distances and their gradients via a euclidean signed distance field. We demonstrate through a comparative study that our approach only slightly increases the computational complexity of the MPC planning. Finally, we validate the effectiveness of our framework through a set of hardware experiments involving dynamic mobile manipulation tasks with potential collisions, such as locomotion balancing with the swinging arm, weight throwing, and autonomous door opening.

• The paper presents a novel collision-free Model Predictive Control (MPC) framework for dynamic legged robot locomotion and manipulation.
• It integrates self-collision and environmental collision avoidance into the MPC cost function using penalty methods and efficient collision modeling pipelines.
• The approach is validated through hardware experiments on dynamic tasks like balancing, weight throwing, and navigation, demonstrating real-time applicability with moderate computational overhead.

Collision-Free Model Predictive Control for Dynamic Legged Mobile Manipulation

The paper "A Collision-Free MPC for Whole-Body Dynamic Locomotion and Manipulation" presents an advanced strategy for real-time planning of legged robot locomotion and manipulation tasks, specifically targeting the challenges of self-collision and environmental collision avoidance. The researchers devise a solution within the framework of Model Predictive Control (MPC) by integrating avoidance constraints as soft penalty functions in a multi-contact optimal control problem.

Methodology

1. MPC Framework Enhancements: The proposed approach extends the whole-body MPC framework by integrating self-collision avoidance mechanisms and environmental collision considerations. The self-collision avoidance is enacted by penalizing distances among collision bodies through representative primitive collision bodies, while environmental collisions are handled via efficient queries of Euclidean signed distance fields.
2. Collision Modeling and Distance Pipelines: The work employs both detailed and simplified collision models for the robot body using primitive geometric shapes such as boxes and cylinders. It further automates the sphere approximation for collision models using an algorithmic approach to streamline the processing and adapt to various robot architectures efficiently.
3. Self and Environment Collision Strategies: Multiple strategies for collision avoidance are analyzed, including the naive narrow-phase approach and a broad-phase manager technique that reduces computational complexity by efficiently handling collision detection hierarchies.

Experimental Demonstrations

The efficacy of this framework was demonstrated through hardware experiments in scenarios involving dynamic tasks that are prone to collisions:

• Balancing with Arm: The robot performed dynamic maneuvers such as changing roll angles or sideways trotting without self-collisions between manipulator and torso components.
• Weight Throwing: Manipulating objects such as throwing a weight dynamically, while avoiding collisions, highlighting the framework's adaptability and robustness.
• Navigation: Environment collision avoidance was validated with dynamic scene adaptation, including autonomous door passing and obstacle negotiation.

Computational Efficiency

The paper benchmarks various collision avoidance techniques in terms of computational overhead during the MPC iterations. Findings suggest that incorporating collision avoidance increases computation time moderately, but remains within acceptable bounds for real-time applications, validating the trade-off between safety and computational load.

Implications and Future Directions

This work holds significant implications for enhancing autonomous legged robot applications in complex dynamic environments without requiring offline planning steps. It contributes to the wider field of robotics by integrating robust collision avoidance in real-time MPC frameworks. Future developments may explore higher-fidelity environmental sensing and more intricate manipulator configurations, along with potential applications in fully autonomous industrial settings and unstructured environments.

This enhancement of the MPC framework underlines the importance of safety in autonomous robotic tasks, addressing both theoretical and practical dimensions. The ability to perform complex manipulative and navigational tasks autonomously without collision threats widens the scope of legged robots for industrial, exploration, and rescue scenarios.