An official website of the United States government
Modern nonlinear control theory seeks to endow systems with properties such as stability and safety, and has been deployed successfully across various domains. Despite this success, model uncertainty remains a significant challenge in ensuring that model-based controllers transfer to real world systems. This paper develops a data-driven approach to robust control synthesis in the presence of model uncertainty using Control Certificate Functions (CCFs), resulting in a convex optimization based controller for achieving properties like stability and safety. An important benefit of our framework is nuanced data-dependent guarantees, which in principle can yield sample-efficient data collection approaches that need not fully determine the input-to-state relationship. This work serves as a starting point for addressing important questions at the intersection of nonlinear control theory and non-parametric learning, both theoretical and in application. We demonstrate the efficiency of the proposed method with respect to input data in simulation with an inverted pendulum in multiple experimental settings.
more »
« less
Model Evolutionary Gain-Based Predictive Control (MEGa-PC) for Soft Robotics*
This paper details a reliable control method for highly nonlinear dynamical systems such as soft robots. We call this method model evolutionary gain-based predictive control or MEGa-PC. The method uses an evolutionary algorithm to optimize a set of controller gains via model predictive control. We demonstrate the performance of MEGa-PC in simulation for a single-link inverted pendulum and a threelink inverted pendulum, and on physical hardware for a threejoint continuum soft robot arm with six degrees of freedom. MEGa-PC is compared to prior work that used Nonlinear Evolutionary Model Predictive Control or NEMPC. The new method performs similarly to NEMPC in terms of accumulated cost over the entire trajectory, however, MEGa-PC generalizes better to real-world applications where safety is paramount, the dynamic model is uncertain, the system has significant latency, and where the previous sampling-based method (NEMPC) resulted in significant steady-state error due to model inaccuracy.
more »
« less
- Award ID(s):
- 2024792
- PAR ID:
- 10561968
- Publisher / Repository:
- IEEE
- Date Published:
- ISSN:
- 2769-4534
- ISBN:
- 979-8-3503-8181-8
- Page Range / eLocation ID:
- 816 to 823
- Format(s):
- Medium: X
- Location:
- San Diego, CA, USA
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
A new control paradigm using angular momentum and foot placement as state variables in the linear inverted pendulum model has expanded the realm of possibilities for the control of bipedal robots. This new paradigm, known as the ALIP model, has shown effectiveness in cases where a robot's center of mass height can be assumed to be constant or near constant as well as in cases where there are no non-kinematic restrictions on foot placement. Walking up and down stairs violates both of these assumptions, where center of mass height varies significantly within a step and the geometry of the stairs restrict the effectiveness of foot placement. In this paper, we explore a variation of the ALIP model that allows the length of the virtual pendulum formed by the robot's stance foot and center of mass to follow smooth trajectories during a step. We couple this model with a control strategy constructed from a novel combination of virtual constraint-based control and a model predictive control algorithm to stabilize a stair climbing gait that does not soley rely on foot placement. Simulations on a 20-degree of freedom model of the Cassie biped in the SimMechanics simulation environment show that the controller is able to achieve periodic gait.more » « less
-
The complex dynamics of agile robotic legged locomotion requires motion planning to intelligently adjust footstep locations. Often, bipedal footstep and motion planning use mathematically simple models such as the linear inverted pendulum, instead of dynamically-rich models that do not have closed-form solutions. We propose a real-time optimization method to plan for dynamical models that do not have closed form solutions and experience irrecoverable failure. Our method uses a data-driven approximation of the step-to-step dynamics and of a failure margin function. This failure margin function is an oriented distance function in state-action space where it describes the signed distance to success or failure. The motion planning problem is formed as a nonlinear program with constraints that enforce the approximated forward dynamics and the validity of state-action pairs. For illustration, this method is applied to create a planner for an actuated spring-loaded inverted pendulum model. In an ablation study, the failure margin constraints decreased the number of invalid solutions by between 24 and 47 percentage points across different objectives and horizon lengths. While we demonstrate the method on a canonical model of locomotion, we also discuss how this can be applied to data-driven models and full-order robot models.more » « less
-
Aggressive and accurate control of complex dynamical systems, such as soft robots, is especially challenging due to the difficulty of obtaining an accurate and tractable model for real-time control. Learned dynamic models are incredibly useful because they do not require derivation of an analytical model, they can represent complex, nonlinear behavior directly from data, and they can be evaluated quickly on graphics-processing units (GPUs). In this paper, we present an open-source Python library to further current research in model-based control of soft robot systems. Our library for Modeling of Learned Dynamics (MoLDy), is designed to generate learned forward models of complex systems through a data-driven approach to hyperparameter optimization and learned model training. Included in the MoLDy library, we present an open-source version of NEMPC (Nonlinear Evolutionary Model Predictive Control), a previously published control algorithm validated on soft robots. We demonstrate the ability of MoLDy and NEMPC to accurately perform modelbased control on a physical pneumatic continuum joint. We also present a benchmarking study on the effect of the loss metric used in model training on control performance. The results of this paper serve to guide other researchers in creating learned dynamic models of novel systems and using them in closed-loop control tasks.more » « less
-
Abstract This paper aims to develop a distributed layered control framework for the navigation, planning, and control of multi-agent quadrupedal robots subject to environments with uncertain obstacles and various disturbances. At the highest layer of the proposed layered control, a reference path for all agents is calculated, considering artificial potential fields (APF) under a priori known obstacles. Second, in the middle layer, we employ a distributed nonlinear model predictive control (NMPC) scheme with a one-step delay communication protocol (OSDCP) subject to reduced-order and linear inverted pendulum (LIP) models of agents to ensure the feasibility of the gaits and collision avoidance, addressing the degree of uncertainty in real-time. Finally, low-level nonlinear whole-body controllers (WBCs) impose the full-order locomotion models of agents to track the optimal and reduced-order trajectories. The proposed controller is validated for effectiveness and robustness on up to four A1 quadrupedal robots in simulations and two robots in the experiments.1 Simulations and experimental validations demonstrate that the proposed approach can effectively address the real-time planning and control problem. In particular, multiple A1 robots are shown to navigate various environments, maintaining collision-free distances while being subject to unknown external disturbances such as pushes and rough terrain.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/RoboSoft60065.2024.10521954
