Safe Output Feedback Motion Planning from Images via Learned Perception Modules and Contraction Theory

Chou, Glen; Özay, Necmiye; Berenson, Dmitry

doi:10.1007/978-3-031-21090-7_21

Cited by 10 publications

(2 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, a vast amount of strategies leverage on the structure of nonlinear control-affine systems to accomplish trajectorytracking. These strategies range from geometric mechanics and geometric nonlinear control theory [9], model-based and reinforcement learning [10], event triggered [11], decentralised robust output feedback [12], neural and deep neural networks [13], iterative learning and data-driven techniques [14], [15], adaptive and tube-based control [16], motion planning leveraging on computer vision and learned perception [17] and Lie algebraic notions [18].…”

Section: A Trajectory-tracking Of Nonlinear Control-affine Systemsmentioning

confidence: 99%

Optimisation-based Quasi Algebraic Controller for trajectory-tracking of a Class of Affine Systems with Constrained Controls

Deniz,

Cheein

2024

Preprint

View full text Add to dashboard Cite

Controlling the locomotion of autonomous legged robots necessitates sophisticated techniques to address system constraints, imposing significant computational demands that hinder embedded implementation. This paper proposes a novel trajectory-tracking approach for a class of affine systems with constrained controls, tailored specifically for autonomous navigation. Leveraging an algebraic strategy initially designed for trajectory-tracking in unmanned wheeled vehicles, our methodology reformulates the approach into a streamlined optimisation problem. This reformulation ensures resulting controls adhere to the robot's actuator limits and accelerations, facilitating smoother, more energy-efficient manoeuvres. The theoretical stability of the proposed controller is established, and empirical validation is conducted through field experiments. Comparative analysis against a nonlinear model predictive controller (iNMPC) with integral mode demonstrates comparable performance, with significantly reduced computational overhead. Specifically, our controller achieves results akin to an iNMPC with a control horizon of 1 (s), sampled every 0.05 (s), showcasing its efficiency in autonomous legged robot navigation.

show abstract

Section: A Trajectory-tracking Of Nonlinear Control-affine Systemsmentioning

confidence: 99%

Optimisation-based Quasi Algebraic Controller for trajectory-tracking of a Class of Affine Systems with Constrained Controls

Deniz,

Cheein

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The state estimator X(y) and the error bound ∆ X (y) may be obtained using machine learning methods, see e.g., [15], [58], or X(y) can encode the extended Kalman filter together with ∆ X (y), see e.g., [59], [60]. We now define the set of admissible inverse output measurement maps as…”

Section: Assumptionmentioning

confidence: 99%

Learning Robust Output Control Barrier Functions From Safe Expert Demonstrations

Lindemann,

Robey,

Jiang

et al. 2024

IEEE Open J. Control. Syst.

View full text Add to dashboard Cite

This paper addresses learning safe output feedback control laws from partial observations of expert demonstrations. We assume that a model of the system dynamics and a state estimator are available along with corresponding error bounds, e.g., estimated from data in practice. We first propose robust output control barrier functions (ROCBFs) as a means to guarantee safety, as defined through controlled forward invariance of a safe set. We then formulate an optimization problem to learn ROCBFs from expert demonstrations that exhibit safe system behavior, e.g., data collected from a human operator or an expert controller. When the parametrization of the ROCBF is linear, then we show that, under mild assumptions, the optimization problem is convex. Along with the optimization problem, we provide verifiable conditions in terms of the density of the data, smoothness of the system model and state estimator, and the size of the error bounds that guarantee validity of the obtained ROCBF. Towards obtaining a practical control algorithm, we propose an algorithmic implementation of our theoretical framework that accounts for assumptions made in our framework in practice. We validate our algorithm in the autonomous driving simulator CARLA and demonstrate how to learn safe control laws from simulated RGB camera images. Control Barrier Functions, Data-driven Robust Control, Output Feedback Control. INDEX TERMS

show abstract