A Unified and Modular Model Predictive Control Framework for Soft Continuum Manipulators under Internal and External Constraints

Filippo A. Spinelli, Robert K. Katzschmann

TL;DR

This paper presents a novel, unified, and modular Model Predictive Control (MPC) framework for controlling soft continuum robots. By adapting MPC to a soft robotic arm (SoPrA) and operating in task-space, the authors demonstrate effective control that handles internal actuation limits, external workspace constraints, and dynamic motions, a significant step towards more capable and versatile soft robotics.

What Problem Does This Paper Address?

Soft continuum robots, with their inherent compliance and safety, hold immense promise for applications ranging from delicate manipulation to human-robot interaction. However, controlling these robots is notoriously challenging. Unlike rigid robots, soft robots exhibit complex, nonlinear actuation dynamics, variable stiffness, and significant shape constraints. Existing control methods often struggle to simultaneously address these issues, typically focusing on joint-space control or handling only a subset of constraints. This leaves a gap in developing a unified and robust control strategy that can manage the full spectrum of challenges, enabling dynamic motions and complex tasks in real-world scenarios.

Methods

The core of this work is the adaptation of Model Predictive Control (MPC), a powerful optimization-based control technique, to the domain of soft robotics. The authors leverage the Augmented Rigid Body Model and the Piece-wise Constant Curvature (PCC) assumption to represent the robot's dynamics. A key innovation is the shift from joint-space to task-space MPC, allowing for more intuitive control of the end-effector's position and orientation.

The framework is designed to be modular, enabling the integration of various constraints:

  • Internal Constraints: Actuation limits (e.g., pressure ranges) and their rate of change are incorporated.
  • External Constraints: Workspace limitations and obstacle avoidance are handled.
  • Dynamic Model: The controller accounts for the robot's inertia, Coriolis forces, gravity, and stiffness/damping.

To handle model uncertainties and disturbances, a Robust Tube-based MPC approach is employed. The authors also explore two methods for incorporating external constraints: a penalized MPC approach that adds a cost to states near obstacles, and a Soft-MPC approach that pre-computes a conservative set of admissible configurations offline.

Datasets

Not discussed in the paper. The authors primarily rely on simulations using a custom URDF model and real-world experiments with their SoPrA robot.

Key Insights and Ideas

  • Task-Space MPC for Soft Robots: Demonstrating that task-space MPC, often used for rigid robots, is not only feasible but highly effective for dynamic control of soft continuum manipulators. This offers a more intuitive control paradigm.
  • Modular Framework: The proposed framework's modularity is a significant advantage, allowing for the seamless integration of various internal and external constraints, as well as future extensions (e.g., incorporating machine learning techniques).
  • Unified Constraint Handling: The ability to handle actuation limits, workspace boundaries, and obstacle avoidance within a single MPC optimization problem is a crucial step towards robust soft robot control.
  • Robustness to Uncertainty: The use of Robust Tube-based MPC effectively addresses model uncertainties and disturbances, leading to more reliable performance in real-world scenarios.
  • Trade-offs in Constraint Handling: The comparison between penalized MPC and Soft-MPC highlights interesting trade-offs. Penalized MPC offers more flexibility online but can be less conservative, while Soft-MPC provides stronger guarantees but can be overly restrictive offline, especially for sparse constraints.

Key Results

  • Simulation and Experimental Validation: The framework was validated through extensive simulations and real-world experiments on the SoPrA robot.
  • Accurate Trajectory Tracking: The MPC controller achieved accurate trajectory tracking for circular motions, outperforming a quasi-static controller in terms of speed and responsiveness.
  • Effective Obstacle Avoidance: Both the penalized MPC and Soft-MPC approaches demonstrated the ability to avoid obstacles, with the penalized MPC adapting more dynamically to the presence of constraints.
  • Handling Unreachable References: The MPC's ability to minimize error even when faced with unreachable references, without needing to reduce the error to zero (as might happen with integral control), is a key advantage for applications requiring operation at the limits of the robot's capabilities.
  • Improved Corner Handling: The future knowledge inherent in MPC allows for better handling of abrupt changes in reference trajectories (corners) without overshooting.

Limitations and Future Work

  • Computational Cost: The MPC formulation, especially with robust elements and offline constraint computation, can be computationally intensive. The current implementation runs at 15 Hz, with potential for improvement to 30-50 Hz through multi-threading and more efficient solvers.
  • Model Approximations: While the Augmented Rigid Body Model and PCC assumption are effective, they are still approximations. The authors acknowledge that further improvements could come from data-driven models or more sophisticated kinematic representations.
  • Offline Constraint Computation: The Soft-MPC approach, which relies on offline computation of constraint sets, can be overly conservative for sparse obstacles.
  • Actuation Delay: The current model does not explicitly account for pressure actuation delay, which was observed in experiments. Future work could incorporate this using augmented state-space models.
  • Future Directions: The authors suggest exploring more advanced solvers, integrating machine learning for model enrichment, increasing the planning horizon for better energy utilization, and further investigating control of underactuated soft robots.

Summary updated

Abstract

Fluidically actuated soft robots have promising capabilities such as inherent compliance and user safety. The control of soft robots needs to properly handle nonlinear actuation dynamics, motion constraints, workspace limitations, and variable shape stiffness, so having a unique algorithm for all these issues would be extremely beneficial. In this work, we adapt Model Predictive Control (MPC), popular for rigid robots, to a soft robotic arm called SoPrA. We address the challenges that current control methods are facing, by proposing a frame-work that handles these in a modular manner. While previous work focused on Joint-Space formulations, we show through simulation and experimental results that Task-Space MPC can be successfully implemented for dynamic soft robotic control. We provide a way to couple the Piece-wise Constant Curvature and Augmented Rigid Body Model assumptions with internal and external constraints and actuation dynamics, delivering an algorithm that unites these aspects and optimizes over them. We believe that a MPC implementation based on our approach could be the way to address most of model-based soft robotics control issues within a unified and modular framework, while allowing to include improvements that usually belong to other control domains such as machine learning techniques.
BibTeX
@inproceedings{spinelli2022unified,
  author = {Filippo A. Spinelli and Robert K. Katzschmann},
  title = {{A Unified and Modular Model Predictive Control Framework for Soft Continuum Manipulators under Internal and External Constraints}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2022},
  url = {https://openreview.net/forum?id=SXR1iGMiIJ},
}
Figure