Fully Autonomous Socially Assistive Robotics: Imitation Learning in a Data Poor Environment
Open Access DepositedImitation learning is a powerful technique for teaching autonomous agents a wide variety of functional tasks. While many advancements have been made in the field of imitation learning, there are very few studies addressing the issues of data limitations. In particular, very few studies address whether imitation learning algorithms can perform well in environments where rich datasets, advanced camera setups, or high-level computational resources are not available. This study aims to address these concerns by exploring imitation learning in data poor environments. Particularly, this study aims to explore whether or not imitation learning can be applied to the development of mobile socially assistive robotics, a domain where having large amounts of high-quality data is difficult to achieve. In this study, two imitation learning algorithms are explored. Namely, a custom multi-branched neural network architecture and Diffusion Policy, an algorithm developed by researchers at Columbia University. Through this study, it is shown that even when operating in data poor environments, imitation learning for socially assistive robotic tasks is not only feasible but achievable. Using generative learning algorithms, like Diffusion Policy, our study demonstrates that imitation learning in data-constrained environments can effectively train robots to perform complex tasks, leveraging minimal datasets to produce robust and adaptable behavioral models. Along with the results from this study, future planned studies are introduced for further exploration.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.