Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Achieving Safe Control Online through Integration of Harmonic Control Lyapunov-Barrier Functions with Unsafe Object-Centric Action Policies

Marlow Fawn, Matthias Scheutz

TL;DR

This work tackles safe online adaptation of pretrained robot policies in dynamic environments by fusing Signal Temporal Logic (STL) constraints with Harmonic Control Lyapunov–Barrier Functions (HCLBFs). The method derives HCLBFs from a restricted STL fragment and solves a Laplace-based potential field on a grid to produce a safety gradient that guides an existing policy via a shared velocity command, with a safety filter that minimally alters actions to ensure safety. A proof-of-concept demonstrates an object-centric force-based policy learned via Soft Actor-Critic (SAC) and deployed on a planar robot arm, safeguarded to avoid obstacles while pursuing a goal. The approach provides formal safety guarantees, supports online adaptation, and can be extended to richer temporal specifications and higher-dimensional workspaces, offering a pathway to dynamically safe behavior without retraining the policy.

Abstract

We propose a method for combining Harmonic Control Lyapunov-Barrier Functions (HCLBFs) derived from Signal Temporal Logic (STL) specifications with any given robot policy to turn an unsafe policy into a safe one with formal guarantees. The two components are combined via HCLBF-derived safety certificates, thus producing commands that preserve both safety and task-driven behavior. We demonstrate with a simple proof-of-concept implementation for an object-centric force-based policy trained through reinforcement learning for a movement task of a stationary robot arm that is able to avoid colliding with obstacles on a table top after combining the policy with the safety constraints. The proposed method can be generalized to more complex specifications and dynamic task settings.

Achieving Safe Control Online through Integration of Harmonic Control Lyapunov-Barrier Functions with Unsafe Object-Centric Action Policies

TL;DR

This work tackles safe online adaptation of pretrained robot policies in dynamic environments by fusing Signal Temporal Logic (STL) constraints with Harmonic Control Lyapunov–Barrier Functions (HCLBFs). The method derives HCLBFs from a restricted STL fragment and solves a Laplace-based potential field on a grid to produce a safety gradient that guides an existing policy via a shared velocity command, with a safety filter that minimally alters actions to ensure safety. A proof-of-concept demonstrates an object-centric force-based policy learned via Soft Actor-Critic (SAC) and deployed on a planar robot arm, safeguarded to avoid obstacles while pursuing a goal. The approach provides formal safety guarantees, supports online adaptation, and can be extended to richer temporal specifications and higher-dimensional workspaces, offering a pathway to dynamically safe behavior without retraining the policy.

Abstract

We propose a method for combining Harmonic Control Lyapunov-Barrier Functions (HCLBFs) derived from Signal Temporal Logic (STL) specifications with any given robot policy to turn an unsafe policy into a safe one with formal guarantees. The two components are combined via HCLBF-derived safety certificates, thus producing commands that preserve both safety and task-driven behavior. We demonstrate with a simple proof-of-concept implementation for an object-centric force-based policy trained through reinforcement learning for a movement task of a stationary robot arm that is able to avoid colliding with obstacles on a table top after combining the policy with the safety constraints. The proposed method can be generalized to more complex specifications and dynamic task settings.

Paper Structure

This paper contains 15 sections, 17 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Signal Temporal Logic (STL)
  4. Harmonic Control Lyapunov–Barrier Functions (HCLBFs)
  5. Reinforcement Learning
  6. STL-Based HCLBFs
  7. STL Definitions
  8. HCLBF Definitions
  9. Laplace Solution for HCLBFs
  10. Proof-of-concept Setting
  11. Policy Learning
  12. Robot Controller
  13. Control Synthesis
  14. Discussion
  15. Conclusion

Figures (2)

  • Figure 1: HCLBF field and gradient (left) and sample trajectories demonstrating safe goal reaching (right).
  • Figure 2: An image from the implementation of the above method in the RoboSuite simulation environment showing the task arena with obstacles (red) and the robot arm's goal point (green).