A Unified MPC Framework for Whole-Body Dynamic Locomotion and Manipulation (2103.00946v1)

Abstract: In this paper, we propose a whole-body planning framework that unifies dynamic locomotion and manipulation tasks by formulating a single multi-contact optimal control problem. We model the hybrid nature of a generic multi-limbed mobile manipulator as a switched system, and introduce a set of constraints that can encode any pre-defined gait sequence or manipulation schedule in the formulation. Since the system is designed to actively manipulate its environment, the equations of motion are composed by augmenting the robot's centroidal dynamics with the manipulated-object dynamics. This allows us to describe any high-level task in the same cost/constraint function. The resulting planning framework could be solved on the robot's onboard computer in real-time within a model predictive control scheme. This is demonstrated in a set of real hardware experiments done in free-motion, such as base or end-effector pose tracking, and while pushing/pulling a heavy resistive door. Robustness against model mismatches and external disturbances is also verified during these test cases.

• The paper introduces a unified control problem that integrates dynamic locomotion and manipulation using a switched-systems perspective.
• It augments the dynamic model by incorporating object properties, enhancing task execution accuracy and operational efficiency.
• Real-time MPC implementation is demonstrated on multi-limbed platforms, effectively managing disturbances and complex constraints.

A Unified MPC Framework for Whole-Body Dynamic Locomotion and Manipulation

The paper presents a novel whole-body planning framework that integrates dynamic locomotion and manipulation tasks using a unified multi-contact optimal control problem. The authors propose a comprehensive model for multi-limbed mobile manipulators by treating them as switched systems, allowing the design of constraints that accommodate both predefined gait sequences and manipulation schedules. The framework combines the robot's centroidal dynamics with the dynamics of the manipulated object, facilitating the description of high-level tasks within a single cost and constraint function.

This framework represents a significant advancement in the field of robotic systems. It employs a nonlinear model predictive control (MPC) approach based on the Sequential-Linear-Quadratic technique, a continuous-time variant of the Iterative-Linear-Quadratic method suited for solving complex optimal control problems. This method has been enhanced to handle equality and inequality constraints, enabling the formulation of a sophisticated optimal control problem that incorporates the dynamic properties of the robot and objects being manipulated.

Several contributions stand out in this research:

1. Unified Control Problem: The integration of locomotion and manipulation into a single optimal control problem using a switched-systems perspective simplifies the formulation of multi-contact tasks. The inclusion of constraints to manage gait sequences and contact schedules adds flexibility and robustness to the control framework.
2. Augmented Dynamic Model: The robot's dynamics are expanded to incorporate the manipulated object's properties, resulting in a comprehensive model that improves task execution accuracy and efficiency. This is particularly relevant for tasks requiring coordinated whole-body maneuvers.
3. Real-Time Implementation: The MPC framework can be deployed in real-time due to its computational efficiency. The authors demonstrate its feasibility through experiments on a multi-limbed robotic platform, executing various dynamic tasks like base or end-effector pose tracking and the manipulation of heavy objects such as pushing and pulling a heavy resistive door.

The experimental results highlight the robustness of this framework, showcasing its capability to handle model mismatches and external disturbances effectively. The implications of this research extend to the practical domains of robotics where seamlessly integrating locomotion and manipulation can lead to more versatile and autonomous robotic systems.

Looking forward, this framework opens new avenues for research in robotic control systems, particularly in developing adaptive and intelligent systems capable of learning and refining their control strategies through interaction with dynamic environments. Future work could focus on integrating collision avoidance capabilities and improving adaptability to varying object parameters, potentially utilizing machine learning techniques to enhance predictive capabilities.

This research represents a cohesive step towards more intelligent and adaptable robotic systems, poised to make significant impacts in industrial automation, rescue operations, and service robotics, where complex whole-body coordination is imperative.