Control Barrier Functions for Feasible Optimization Under Actuation Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optimization-based controllers for autonomous systems fail to effectively manage the feasibility of user constraints under actuation limits, particularly due to the dynamic nature of system states, leading to potential infeasibility without considering the volume and rate of change of the feasible space.
Innovation Solution
The proposed method monitors and modifies user constraints by augmenting them with new constraints based on the volume and gradient of the feasible space, ensuring the feasible space volume is maintained within bounds, thereby preventing or delaying infeasibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If user constraints are strictly enforced in optimization-based control, then control safety is improved, but feasible space volume decreases leading to potential infeasibility
Solution Approach 1:
The patent applies dynamics by making the control barrier function parameters adaptive rather than fixed. The parameters are updated in real-time based on the system state and the volume of feasible space, allowing the constraints to dynamically adjust their stringency. This resolves the contradiction by enabling safety enforcement that adapts to the current feasible space conditions, preventing infeasibility while maintaining safety where possible.
Solution Approach 2:
The patent changes the parameters of the control barrier functions based on the volume of feasible space and system state. By modifying these parameters adaptively, the optimization problem maintains feasibility while still enforcing safety constraints. This directly addresses the contradiction by allowing the constraint parameters to change in response to the feasible space volume, preventing the feasible space from becoming empty.
2Reliability
If control barrier function parameters are updated frequently to maintain feasibility, then continuous feasibility is improved, but computational complexity increases
Solution Approach 1:
The patent computes the volume of feasible space and determines parameter updates in advance of when infeasibility would occur. By proactively adjusting parameters based on the current feasible space volume, the system prevents infeasibility rather than reacting to it, ensuring continuous feasibility while managing computational complexity through planned rather than reactive updates.
Solution Approach 2:
The patent implements feedback by continuously monitoring the volume of feasible space and using this information to adjust control barrier function parameters. This closed-loop approach ensures that parameters are updated based on actual system conditions, maintaining continuous feasibility while avoiding unnecessary updates that would increase computational complexity.
3Stability of the object's composition
If feasible space volume is restricted to prevent infeasibility, then controller stability is improved, but control flexibility decreases
Solution Approach 1:
The patent applies dynamics by making the feasible space restrictions adaptive rather than static. The restrictions on feasible space volume are adjusted in real-time based on system state and safety requirements, allowing the controller to maintain stability while adapting its flexibility to current conditions. This resolves the contradiction by enabling stability through controlled restrictions that can relax when safe.
Data Source
AI summary
A method, computer program product, and computer system for obtaining, by a computing device, a system state and dynamics model. User constraints in a controller for a control optimization may be obtained. A feasible space volume may be determined based upon, at least in part, the system state and dynamics model. A new control optimization may be determined based upon, at least in part, the feasible space volume. A control input for the controller may be executed based upon, at least in part, the new control optimization.