Prediction and Control of Stochastic Agents Using Formal Methods

6 pages•Published: October 23, 2023

Avraham Raviv, Yuval Gerber, Liri Benzinou, Michelle Aluf-Medina and Hillel Kugler

Abstract

In this paper, we propose an innovative approach that incorporates formal verification methods into the training process of stochastic Reinforcement Learning (RL) agents. Our method allows for the analysis and improvement of the learning process of these agents. Specifically, we demonstrate the capability to evaluate RL policies (prediction) and opti- mize them (control) using different model checkers. The integration of formal verification tools with stochastic RL agents strengthens the applicability and potential of our approach, paving the way for more robust and reliable learning systems.

Keyphrases: formal verification, model checking., reinforcement learning

In: Nina Narodytska, Guy Amir, Guy Katz and Omri Isac (editors). Proceedings of the 6th Workshop on Formal Methods for ML-Enabled Autonomous Systems, vol 16, pages 29-34.

Links:	https://easychair.org/publications/paper/hWbP
	https://doi.org/10.29007/1q69

BibTeX entry

@inproceedings{FoMLAS2023:Prediction_Control_Stochastic_Agents,
  author    = {Avraham Raviv and Yuval Gerber and Liri Benzinou and Michelle Aluf-Medina and Hillel Kugler},
  title     = {Prediction and Control of Stochastic Agents Using Formal Methods},
  booktitle = {Proceedings of the 6th Workshop on Formal Methods for ML-Enabled Autonomous Systems},
  editor    = {Nina Narodytska and Guy Amir and Guy Katz and Omri Isac},
  series    = {Kalpa Publications in Computing},
  volume    = {16},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2515-1762},
  url       = {/publications/paper/hWbP},
  doi       = {10.29007/1q69},
  pages     = {29-34},
  year      = {2023}}

Download PDF Open PDF in browser