Experience-driven Predictive Control with Robust Constraint Satisfaction under Time-Varying State Uncertainty

Abstract

Abstract-We present an extension to Experience-driven Predictive Control (EPC) that leverages a Gaussian belief propagation strategy to compute an uncertainty set bounding the evolution of the system state in the presence of time-varying state uncertainty. This uncertainty set is used to tighten the constraints in the predictive control formulation via a chance constrained approach, thereby providing a probabilistic guarantee of constraint satisfaction. The parameterized form of the controllers produced by EPC coupled with online uncertainty estimates ensures this robust constraint satisfaction property persists even as the system switches controllers and experiences variations in the uncertainty model. We validate the online performance and robust constraint satisfaction of the proposed Robust EPC algorithm through a series of experimental trials with a small quadrotor platform subjected to changes in state estimate quality.

Leveraging experience for robust, adaptive nonlinear MPC on computationally constrained systems with time-varying state uncertainty

“…This manuscript refines an earlier conference presentation of the Robust EPC algorithm (Desaraju et al, 2017), including a more detailed discussion of the model adaptation techniques in Section 2.5. We also present new simulation results with a ground robot (Section 3.1), and several new experimental studies with three different platforms: a skid-steer ground robot equipped with a laser-based localization system (Section 3.3), a hexarotor aerial robot equipped with a vision-based localization system (Section 3.4), and additional results with a small quadrotor aerial robot executing aggressive maneuvers in a strong, spatially varying wind field (Section 3.2.4).…”

Learning-Based Model Predictive Control: Toward Safe Learning in Control

“…Due to many technical challenges in stochastic MPC and some of the issues discussed above, many stochastic learning-based MPC schemes come with little theoretical analysis, but they have nevertheless been very successful in practical implementations. For instance, Desaraju and colleagues (77,78) presented parametric learning-based MPC approaches for robotic systems that enhance the model of a quadrotor over time while taking approximate uncertainty in the prediction into account through constraint tightening. Similarly, McKinnon & Schoellig (79) presented an implementation for trajectory tracking of an off-road robotic vehicle based on linear Bayesian regression for the actuator dynamics, with the particular focus of mitigating the conflict between fast adaptation and long term learning.…”

Learn Fast, Forget Slow: Safe Predictive Learning Control for Systems With Unknown and Changing Dynamics Performing Repetitive Tasks

“…This means that all data gathered by the robot can be grouped into one model and used to train any model parameters and validate them to avoid overfitting. This class of methods has shown impressive results control of ground robots [2], [7], quadrotors [6], manipulators [5] and humanoid robots [8]. This style of approach can learn new dynamics quickly, but if the robot dynamics can change due to a factor that is not included in the model (e.g.…”

“…In particular, probabilistic models are used since they provide a measure of model uncertainty which can naturally be used to derive an upper bound on model error. Two common methods for doing this are GP regression [1]- [3] and various forms of local linear regression [4]- [6]. The Fig.…”