Abstract
Reinforcement Learning (RL) depends critically on how reward functions are designed to capture intended behavior. However, traditional approaches are unable to represent temporal behavior, such as “do task 1 before doing task 2.” In the event they can represent temporal behavior, these reward functions are handcrafted by researchers and often require long hours of trial and error to shape the reward function just right to get the desired behavior. In these cases, the desired behavior is already known, the problem is generating a reward function to train the RL agent to satisfy that behavior. To address this issue, we present our approach for automatically converting timed and untimed specifications into a reward function, which has been implemented as the tool STLGym. In this work, we show how STLGym can be used to train RL agents to satisfy specifications better than traditional approaches and to refine learned behavior to better match the specification.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The environment is based on the classic cart-pole system implemented for [3], where more information on the dynamics can be found.
- 2.
STLGym implementation is available at https://github.com/nphamilton/stl-gym.
- 3.
The RTAMT code is available at https://github.com/nickovic/rtamt.
- 4.
If a weight is not defined by the user, the default is 1.
- 5.
All training scripts are available at https://github.com/nphamilton/spinningup/tree/master/spinup/examples/sefm2022.
- 6.
These specifications came from [2].
References
Aksaray, D., Jones, A., Kong, Z., Schwager, M., Belta, C.: Q-learning for robust satisfaction of signal temporal logic specifications. In: 2016 IEEE 55th Conference on Decision and Control (CDC), pp. 6565–6570. IEEE (2016)
Balakrishnan, A., Deshmukh, J.V.: Structured reward shaping using signal temporal logic specifications. In: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3481–3486. IEEE (2019)
Barto, A.G., Sutton, R.S., Anderson, C.W.: Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Trans. Syst. Man Cybern. SMC- 13(5), 834–846 (1983)
Brockman, G., et al.: Openai gym (2016)
Camacho, A., Icarte, R.T., Klassen, T.Q., Valenzano, R.A., McIlraith, S.A.: LTL and beyond: formal languages for reward function specification in reinforcement learning. In: IJCAI. vol. 19, pp. 6065–6073 (2019)
Fujimoto, S., Hoof, H., Meger, D.: Addressing function approximation error in actor-critic methods. In: International Conference on Machine Learning, pp. 1587–1596. PMLR (2018)
Haarnoja, T., Zhou, A., Abbeel, P., Levine, S.: Soft actor-critic: off-policy maximum entropy deep reinforcement learning with a stochastic actor. In: International Conference on Machine Learning, pp. 1861–1870. PMLR (2018)
Hamilton, N., Musau, P., Lopez, D.M., Johnson, T.T.: Zero-shot policy transfer in autonomous racing: reinforcement learning vs imitation learning. In: Proceedings of the 1st IEEE International Conference on Assured Autonomy (2022)
Hamilton, N., Schlemmer, L., Menart, C., Waddington, C., Jenkins, T., Johnson, T.T.: Sonic to knuckles: evaluations on transfer reinforcement learning. In: Unmanned Systems Technology XXII. vol. 11425, p. 114250J. International Society for Optics and Photonics (2020)
Hasanbeig, M., Abate, A., Kroening, D.: Logically-constrained reinforcement learning code repository. https://github.com/grockious/lcrl (2020)
Icarte, R.T., Klassen, T., Valenzano, R., McIlraith, S.: Using reward machines for high-level task specification and decomposition in reinforcement learning. In: International Conference on Machine Learning, pp. 2107–2116. PMLR (2018)
Jothimurugan, K., Alur, R., Bastani, O.: A composable specification language for reinforcement learning tasks. In: Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019 (2019)
Jothimurugan, K., Bastani, O., Alur, R.: Abstract value iteration for hierarchical reinforcement learning. In: International Conference on Artificial Intelligence and Statistics, pp. 1162–1170. PMLR (2021)
Li, X., Vasile, C.I., Belta, C.: Reinforcement learning with temporal logic rewards. In: 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3834–3839. IEEE (2017)
Lillicrap, T.P., et al.: Continuous control with deep reinforcement learning. In: ICLR (2016)
Maler, O., Nickovic, D.: Monitoring temporal properties of continuous signals. In: Lakhnech, Y., Yovine, S. (eds.) FORMATS/FTRTFT -2004. LNCS, vol. 3253, pp. 152–166. Springer, Heidelberg (2004). https://doi.org/10.1007/978-3-540-30206-3_12
Mania, H., Guy, A., Recht, B.: Simple random search of static linear policies is competitive for reinforcement learning. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems, pp. 1805–1814 (2018)
Mnih, V., et al.: Human-level control through deep reinforcement learning. Nature 518(7540), 529–533 (2015)
Ničković, D., Yamaguchi, T.: RTAMT: online robustness monitors from STL. In: Hung, D.V., Sokolsky, O. (eds.) ATVA 2020. LNCS, vol. 12302, pp. 564–571. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-59152-6_34
Peters, J., Mulling, K., Altun, Y.: Relative entropy policy search. In: Twenty-Fourth AAAI Conference on Artificial Intelligence (2010)
Riedmiller, M.: Neural fitted Q iteration – first experiences with a data efficient neural reinforcement learning method. In: Gama, J., Camacho, R., Brazdil, P.B., Jorge, A.M., Torgo, L. (eds.) ECML 2005. LNCS (LNAI), vol. 3720, pp. 317–328. Springer, Heidelberg (2005). https://doi.org/10.1007/11564096_32
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., Klimov, O.: Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347 (2017)
Silver, D., et al.: Mastering the game of go with deep neural networks and tree search. Nature 529, 484–489 (2016)
Silver, D., Singh, S., Precup, D., Sutton, R.S.: Reward is enough. Artif. Intell. 299, 103535 (2021)
Vinyals, O., et al.: Grandmaster level in StarCraft ii using multi-agent reinforcement learning. Nature 575(7782), 350–354 (2019)
Wang, X., Chen, Y., Zhu, W.: A survey on curriculum learning. IEEE Trans. Pattern Anal. Mach. Intell. 44, 4555–4576 (2021)
Watkins, C.J., Dayan, P.: Q-learning. Mach. Learn. 8(3), 279–292 (1992)
Wurman, P.R., et al.: Outracing champion Gran Turismo drivers with deep reinforcement learning. Nature 602(7896), 223–228 (2022)
Acknowledgments
The material presented in this paper is based upon work supported the Defense Advanced Research Projects Agency (DARPA) through contract number FA8750-18-C-0089, the Air Force Office of Scientific Research (AFOSR) award FA9550-22-1-0019, the National Science Foundation (NSF) through grant number 2028001, and the Department of Defense (DoD) through the National Defense Science & Engineering Graduate (NDSEG) Fellowship Program. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of DARPA, AFOSR, NSF or DoD.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Hamilton, N., Robinette, P.K., Johnson, T.T. (2022). Training Agents to Satisfy Timed and Untimed Signal Temporal Logic Specifications with Reinforcement Learning. In: Schlingloff, BH., Chai, M. (eds) Software Engineering and Formal Methods. SEFM 2022. Lecture Notes in Computer Science, vol 13550. Springer, Cham. https://doi.org/10.1007/978-3-031-17108-6_12
Download citation
DOI: https://doi.org/10.1007/978-3-031-17108-6_12
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-17107-9
Online ISBN: 978-3-031-17108-6
eBook Packages: Computer ScienceComputer Science (R0)