Abstract
The ability to learn compositional strategies in multi-task learning and to exert them appropriately is crucial to the development of artificial intelligence. However, there exist several challenges: (i) how to maintain the independence of modules in learning their own sub-tasks; (ii) how to avoid performance degradation in situations where modules’ reward scales are incompatible; (iii) how to find the optimal composite policy for the entire set of tasks. In this paper, we introduce a Modular Reinforcement Learning (MRL) framework that coordinates the competition and the cooperation between separate modules. Furthermore, a selective update mechanism enables the learning system to align incomparable reward scales in different modules. Moreover, the learning system follows a “joint policy” to calculate actions’ preferences combined with their responsibility for the current task. We evaluate the effectiveness of our approach on a classic food-gathering and predator-avoidance task. Results show that our approach has better performance than previous MRL methods in learning separate strategies for sub-tasks, is robust to modules with incomparable reward scales, and maintains the independence of the learning in each module.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Bernard, J.A.: Don’t forget the little brain: a framework for incorporating the cerebellum into the understanding of cognitive aging. Neurosci. Biobehav. Rev. 137, 104639 (2022)
Botvinick, M.M.: Hierarchical models of behavior and prefrontal function. Trends Cogn. Sci. 12(5), 201–208 (2008)
Doya, K., Samejima, K., Katagiri, K.I., Kawato, M.: Multiple model-based reinforcement learning. Neural Comput. 14(6), 1347–1369 (2002)
Esteban, D., Rozo, L., Caldwell, D.G.: Hierarchical reinforcement learning for concurrent discovery of compound and composable policies. In: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1818–1825. IEEE (2019)
Gatti, D., Rinaldi, L., Ferreri, L., Vecchi, T.: The human cerebellum as a hub of the predictive brain. Brain Sci. 11(11), 1492 (2021)
Gupta, V., Anand, D., Paruchuri, P., Kumar, A.: Action selection for composable modular deep reinforcement learning. In: Proceedings of the 20th International Conference on Autonomous Agents and MultiAgent Systems, pp. 565–573 (2021)
Jacobs, R.A., Jordan, M.I., Nowlan, S.J., Hinton, G.E.: Adaptive mixtures of local experts. Neural Comput. 3(1), 79–87 (1991)
Logan, G.D., Crump, M.J.: Hierarchical control of cognitive processes: the case for skilled typewriting. In: Psychology of Learning and Motivation, vol. 54, pp. 1–27. Elsevier (2011)
Nagabandi, A., Kahn, G., Fearing, R.S., Levine, S.: Neural network dynamics for model-based deep reinforcement learning with model-free fine-tuning. In: 2018 IEEE International Conference on Robotics and Automation (ICRA), pp. 7559–7566. IEEE (2018)
Narendra, K.S., Balakrishnan, J., Ciliz, M.K.: Adaptation and learning using multiple models, switching, and tuning. IEEE Control Syst. Mag. 15(3), 37–51 (1995)
Nowlan, S.J., Hinton, G.E.: Evaluation of adaptive mixtures of competing experts. In: NIPS, vol. 3, pp. 774–780 (1990)
Samejima, K., Doya, K., Kawato, M.: Inter-module credit assignment in modular reinforcement learning. Neural Netw. 16(7), 985–994 (2003)
Simpkins, C., Isbell, C.: Composable modular reinforcement learning. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, pp. 4975–4982 (2019)
Smith, B.J., Read, S.J.: Modeling incentive salience in Pavlovian learning more parsimoniously using a multiple attribute model. Cogn. Affect. Behav. Neurosci. 22, 244–257 (2021). https://doi.org/10.3758/s13415-021-00953-2
Sodhani, S., Zhang, A., Pineau, J.: Multi-task reinforcement learning with context-based representations. In: International Conference on Machine Learning, pp. 9767–9779. PMLR (2021)
Sprague, N., Ballard, D.: Multiple-goal reinforcement learning with modular Sarsa(0) (2003)
Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction. MIT Press, Cambridge (2018)
Wang, J.X., et al.: Learning to reinforcement learn. arXiv preprint arXiv:1611.05763 (2016)
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
Xue, J., Alexandre, F. (2022). Multi-task Learning with Modular Reinforcement Learning. In: Cañamero, L., Gaussier, P., Wilson, M., Boucenna, S., Cuperlier, N. (eds) From Animals to Animats 16. SAB 2022. Lecture Notes in Computer Science(), vol 13499. Springer, Cham. https://doi.org/10.1007/978-3-031-16770-6_11
Download citation
DOI: https://doi.org/10.1007/978-3-031-16770-6_11
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-16769-0
Online ISBN: 978-3-031-16770-6
eBook Packages: Computer ScienceComputer Science (R0)