HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation

Self-Exploration of Sensorimotor Spaces in Robots

Abstract : Developmental robotics has begun in the last fifteen years to study robots that have a childhood---crawling before trying to run, playing before being useful---and that are basing their decisions upon a lifelong and embodied experience of the real-world. In this context, this thesis studies sensorimotor exploration---the discovery of a robot's own body and proximal environment---during the early developmental stages, when no prior experience of the world is available. Specifically, we investigate how to generate a diversity of effects in an unknown environment. This approach distinguishes itself by its lack of user-defined reward or fitness function, making it especially suited for integration in self-sufficient platforms. In a first part, we motivate our approach, formalize the exploration problem, define quantitative measures to assess performance, and propose an architectural framework to devise algorithms. Through the extensive examination of a multi-joint arm example, we explore some of the fundamental challenges that sensorimotor exploration faces, such as high-dimensionality and sensorimotor redundancy, in particular through a comparison between motor and goal babbling exploration strategies. We propose several algorithms and empirically study their behaviour, investigating the interactions with developmental constraints, external demonstrations and biologically-inspired motor synergies. Furthermore, because even efficient algorithms can provide disastrous performance when their learning abilities do not align with the environment's characteristics, we propose an architecture that can dynamically discriminate among a set of exploration strategies. Even with good algorithms, sensorimotor exploration is still an expensive proposition---a problem since robots inherently face constraints on the amount of data they are able to gather; each observation takes a non-negligible time to collect. In a second part, we propose the reuse algorithm that allows to exploit the exploration trajectories of a previous environment in another new, unknown one, to improve exploration, with the only constraining assumptions being that the two environments share the same motor space---which is often the case as a robot's body remains similar across tasks. No assumption is made that the sensory modalities of the two tasks remain identical, or that the exploration strategies or the learning algorithms are the same. If the latent dynamics of the two environment share some degree of similarity, we establish that the reuse algorithm provides improvements in exploration. We illustrate this on a real robot setup interacting with different objects in augmented reality. We then show that the reuse algorithm can exhibit scaffolding behaviour. This allows to guide skill acquisition through the exclusive manipulation of environments where no reward or fitness function needs to be defined. Additionally, we conduct experiments that show that exploration on real-world robots can benefit from reusing exploration trajectories produced on surrogate, simplified---even purely kinematic---simulations. Throughout this thesis, our core contributions are algorithms description and empirical results. In order to allow unrestricted examination and reproduction of all our results, the entire code is made available. Sensorimotor exploration is a fundamental developmental mechanism of biological systems. By decoupling it from learning and studying it in its own right in this thesis, we engage in an approach that casts light on important problems facing robots developing on their own.
Document type :
Complete list of metadata

Contributor : Fabien C. Y. Benureau Connect in order to contact the contributor
Submitted on : Thursday, January 14, 2016 - 4:04:29 PM
Last modification on : Friday, March 25, 2022 - 3:34:01 PM
Long-term archiving on: : Friday, April 15, 2016 - 4:40:22 PM


  • HAL Id : tel-01251324, version 1



Fabien Benureau. Self-Exploration of Sensorimotor Spaces in Robots. Robotics [cs.RO]. Université de Bordeaux, 2015. English. ⟨tel-01251324v1⟩



Record views


Files downloads