Simultaneous Contact-Rich Grasping and Locomotion via Distributed Optimization Enabling Free-Climbing for Multi-Limbed Robots (2207.01418v2)

Abstract: While motion planning of locomotion for legged robots has shown great success, motion planning for legged robots with dexterous multi-finger grasping is not mature yet. We present an efficient motion planning framework for simultaneously solving locomotion (e.g., centroidal dynamics), grasping (e.g., patch contact), and contact (e.g., gait) problems. To accelerate the planning process, we propose distributed optimization frameworks based on Alternating Direction Methods of Multipliers (ADMM) to solve the original large-scale Mixed-Integer NonLinear Programming (MINLP). The resulting frameworks use Mixed-Integer Quadratic Programming (MIQP) to solve contact and NonLinear Programming (NLP) to solve nonlinear dynamics, which are more computationally tractable and less sensitive to parameters. Also, we explicitly enforce patch contact constraints from limit surfaces with micro-spine grippers. We demonstrate our proposed framework in the hardware experiments, showing that the multi-limbed robot is able to realize various motions including free-climbing at a slope angle 45{\deg} with a much shorter planning time.

• The paper introduces an ADMM-based framework that unifies locomotion, grasping, and contact tasks, enhancing free-climbing capabilities.
• It employs a unified MINLP approach partitioned into MIQP and NLP sub-problems, significantly reducing planning times and improving tractability.
• Experimental validation on a 9.6 kg four-limbed robot on a 45° slope demonstrates stable, efficient trajectories and robust free-climbing performance.

Simultaneous Contact-Rich Grasping and Locomotion via Distributed Optimization Enabling Free-Climbing for Multi-Limbed Robots

The paper presents an optimization-based framework aimed at enhancing the capabilities of multi-limbed robots, particularly in tasks that involve complex interactions with the environment, such as free-climbing. This paper focuses on integrating locomotion and grasping dynamics simultaneously, addressing the unique computational challenges posed by non-linear dynamics, contact constraints, and the need for efficient planning in high-dimensional spaces.

Optimization Framework

The authors propose a novel motion planning framework that unifies the locomotion, grasping, and contact tasks into a single optimization problem. This is achieved by employing an Alternating Direction Method of Multipliers (ADMM) based distributed optimization approach. The framework tackles large-scale Mixed-Integer Nonlinear Programming (MINLP) problems by solving them as a combination of Mixed-Integer Quadratic Programming (MIQP) for contact and Nonlinear Programming (NLP) for dynamic constraints. The ADMM effectively partitions this problem into smaller, more manageable sub-problems, improving tractability and efficiency.

Contact Models with Micro-Spines

A significant contribution of the paper is the exploration of patch contact models using micro-spines, which are essential for multi-finger grasping tasks. These models consider both frictional and spine-based constraints, forming a sophisticated contact model that captures the nuances of physical interaction required for free-climbing. The paper incorporates a limit surface model into the planner, allowing the robot to leverage micro-spine based grip under large shear forces or moments without relying solely on normal force.

Experimental Validation

The proposed algorithm is validated through simulations and hardware experiments on a 9.6 kg four-limbed robot tasked with free-climbing on a rugged terrain marked by a 45-degree slope. This experimental setup showcases the potential for the framework to generate efficient, physically feasible trajectories that account for both locomotion and grasping. Notably, when implementing multi-finger dexterous tasks, the robot demonstrated significantly reduced planning times, supporting the effectiveness of the distributed optimization framework.

Observations and Implications

The experimental results signify reduced trajectory planning times alongside validated stability and contact conditions, promising innovations in robotic applications like planetary exploration and structural inspection, where complex terrain navigation is critical. The robustness of the algorithm, evidenced by its handling of shear forces and dynamic changes, sets a precedent for future work in dynamically rich environments.

Future Developments and Considerations

Future research can focus on enhancing the computational efficiency of the proposed ADMM-based framework. By addressing robustness under variable physical parameters and uncertainties, the work could extend into more adaptive and real-time applications. The exploration of heuristics for online optimization in real-world scenarios could further bolster the practicality of this approach. Enhanced parameter estimation and adaptive control strategies can mitigate discrepancies observed during hardware trials, further solidifying the translation of simulations into reliable robotic maneuvers in the field.

By demonstrating comprehensive planning that is computationally feasible for complex tasks such as free-climbing, this paper contributes significantly to the domain of multi-limbed robotics, paving the way for future research into adaptive and versatile robotic systems.