Litcius/Paper detail

Adaptive Safety with Multiple Barrier Functions Using Integral Concurrent Learning

Axton Isaly, Omkar Sudhir Patil, Ricardo G. Sanfelice, Warren E. Dixon

202132 citationsDOI

Abstract

This paper presents an approach to guarantee safety for control systems with uncertain nonlinear dynamics. Constraints on the control input induced by multiple barrier function candidates are developed to ensure forward pre-invariance of a safe set of states despite the uncertainty. Using the adaptive control technique of integral concurrent learning, conservativeness in the input constraints is reduced over time as estimates of the uncertain parameters exponentially converge. The constraints are implemented in a quadratic program that modifies a nominal controller for guaranteed safety. An example is presented showing that the operating region of the system is expanded significantly from the initial size due to the converging estimation error.

Topics & Concepts

Control theory (sociology)Quadratic equationComputer scienceController (irrigation)Nonlinear systemSet (abstract data type)Function (biology)Mathematical optimizationAdaptive controlControl (management)Quadratic programmingMathematicsArtificial intelligenceEvolutionary biologyAgronomyQuantum mechanicsBiologyGeometryProgramming languagePhysicsAdvanced Control Systems OptimizationAdaptive Control of Nonlinear SystemsControl Systems and Identification
Adaptive Safety with Multiple Barrier Functions Using Integral Concurrent Learning | Litcius