Keep it Upright: Model Predictive Control for Nonprehensile Object Transportation with Obstacle Avoidance on a Mobile Manipulator (2305.17484v2)

Abstract: We consider a nonprehensile manipulation task in which a mobile manipulator must balance objects on its end effector without grasping them -- known as the waiter's problem -- and move to a desired location while avoiding static and dynamic obstacles. In constrast to existing approaches, our focus is on fast online planning in response to new and changing environments. Our main contribution is a whole-body constrained model predictive controller (MPC) for a mobile manipulator that balances objects and avoids collisions. Furthermore, we propose planning using the minimum statically-feasible friction coefficients, which provides robustness to frictional uncertainty and other force disturbances while also substantially reducing the compute time required to update the MPC policy. Simulations and hardware experiments on a velocity-controlled mobile manipulator with up to seven balanced objects, stacked objects, and various obstacles show that our approach can handle a variety of conditions that have not been previously demonstrated, with end effector speeds and accelerations up to 2.0 m/s and 7.9 m/s$^2$, respectively. Notably, we demonstrate a projectile avoidance task in which the robot avoids a thrown ball while balancing a tall bottle.

• The paper presents an innovative MPC framework that enables nonprehensile object transport while proactively avoiding both static and dynamic obstacles.
• It integrates minimum statically-feasible friction coefficients to simplify constraints, enhancing robustness and reducing computational load.
• Experiments confirm the controller's ability to maintain stability at speeds up to 2.0 m/s and accelerations of 7.9 m/s², proving its practical effectiveness.

An Analysis of Model Predictive Control for Nonprehensile Object Transportation

This paper presents a comprehensive paper on employing model predictive control (MPC) for addressing nonprehensile manipulation, specifically focusing on the "waiter's problem." The primary objective is to transport objects balanced on an end effector of a mobile manipulator while effectively avoiding collisions with static and dynamic obstacles. This task is distinctive due to the lack of grasping, instead relying on balancing the objects solely via dynamic interactions governed by unilateral constraints.

The authors detail the development of a whole-body MPC tailored for a velocity-controlled mobile manipulator engaging in dynamic balancing tasks under environmental uncertainty. A significant contribution of the work is the incorporation of minimum statically-feasible friction coefficients within the MPC design. This inclusion offers enhanced robustness against frictional uncertainty and reduces computational demands by simplifying the friction constraints when possible, specifically when these coefficients can be reduced to zero in scenarios involving parallel support planes.

The research underscores the effectiveness of the proposed controller through both simulation and real-world experiments. The authors conduct hardware experiments involving diverse object arrangements, such as individual and stacked objects, under conditions with variable obstacle configurations. The results affirm that the approach can maintain balancing stability and obstacle avoidance at considerable end effector velocities and accelerations, reaching 2.0 m/s and 7.9 m/s², respectively. Furthermore, an innovative scenario demonstrated projectile avoidance, where the manipulator successfully dodged a thrown ball while maintaining balance—a noteworthy testament to the controller’s real-time adaptability and robustness.

The paper situates its contributions within the broader context of related works by contrasting existing solutions that either rely on offline motion planning or sensor feedback mechanisms for balancing. The novel contribution here is the integration of robust online optimization that departs from prior approaches, namely, those that focus on static settings or perceive dynamic disturbances reactively rather than proactively.

Future implications of this research are multifold. Practically, the proposed strategies could be highly relevant in sectors such as warehousing and automated food service, where the need for dexterous and efficient mobile manipulation is paramount. Theoretically, the principles of the robust constraint handling showcased here may influence subsequent work in MPC for robotics, particularly in tasks involving dynamic environments and non-linear system modeling.

Moreover, the findings open avenues for further exploration into scenarios where object parameters, such as inertia and shape, bear uncertainties. The work could potentially extend to adaptive or learning-based frameworks where the MPC adapts its strategies over time or in varying contexts and object configurations.

Overall, the research provides a solid groundwork for advancing nonprehensile manipulation systems, highlighting a balance between control sophistication and computational viability, contributing positively to the growing field of efficient robotic manipulation in dynamic environments.