This work presents an approach for embodied agents that have to learn models from the least amount of prior knowledge, solely based on knowing which actions can be performed and observing the state. Instead of relying on (often black-box) quantitative models, a qualitative forward model is learned that finds the relations among the variables, the contextual relations, and the qualitative influence. We assume qualitative determinism and monotonicity, assumptions motivated by human learning. A learning and exploitation algorithm is designed and demonstrated on a robot with a gripper. The robot can grab an object and move it to another location, without predefined knowledge of how to move, grab or displace objects.
Lemeire, J, Wouters, N, Cleemput, MV & Heirman, A 2024, Contextual Qualitative Deterministic Models for Self-learning Embodied Agents. in CL Buckley, D Cialfi, P Lanillos, M Ramstead, T Verbelen, M Ramstead, N Sajid, H Shimazaki & M Wisse (eds), Active Inference: 4th International Workshop, IWAI 2023, Ghent, Belgium, September 13–15, 2023, Revised Selected Papers. vol. 1915, Communications in Computer and Information Science, vol. 1915 CCIS, Springer, pp. 3-13. https://doi.org/10.1007/978-3-031-47958-8_1
Lemeire, J., Wouters, N., Cleemput, M. V., & Heirman, A. (2024). Contextual Qualitative Deterministic Models for Self-learning Embodied Agents. In C. L. Buckley, D. Cialfi, P. Lanillos, M. Ramstead, T. Verbelen, M. Ramstead, N. Sajid, H. Shimazaki, & M. Wisse (Eds.), Active Inference: 4th International Workshop, IWAI 2023, Ghent, Belgium, September 13–15, 2023, Revised Selected Papers (Vol. 1915, pp. 3-13). (Communications in Computer and Information Science; Vol. 1915 CCIS). Springer. https://doi.org/10.1007/978-3-031-47958-8_1
@inproceedings{318cfcef97bd47f9bb54b5429035198c,
title = "Contextual Qualitative Deterministic Models for Self-learning Embodied Agents",
abstract = "This work presents an approach for embodied agents that have to learn models from the least amount of prior knowledge, solely based on knowing which actions can be performed and observing the state. Instead of relying on (often black-box) quantitative models, a qualitative forward model is learned that finds the relations among the variables, the contextual relations, and the qualitative influence. We assume qualitative determinism and monotonicity, assumptions motivated by human learning. A learning and exploitation algorithm is designed and demonstrated on a robot with a gripper. The robot can grab an object and move it to another location, without predefined knowledge of how to move, grab or displace objects.",
author = "Jan Lemeire and Nick Wouters and Cleemput, \{Marco Van\} and Aron Heirman",
note = "Publisher Copyright: {\textcopyright} 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.",
year = "2024",
doi = "10.1007/978-3-031-47958-8\_1",
language = "English",
isbn = "9783031479571",
volume = "1915",
series = "Communications in Computer and Information Science",
publisher = "Springer",
pages = "3--13",
editor = "Buckley, \{Christopher L.\} and Daniela Cialfi and Pablo Lanillos and Maxwell Ramstead and Tim Verbelen and Maxwell Ramstead and Noor Sajid and Hideaki Shimazaki and Martijn Wisse",
booktitle = "Active Inference",
}