Discovering space — Grounding spatial topology and metric regularity in a naive agent’s sensorimotor experience

Abstract

In line with the sensorimotor contingency theory, we investigate the problem of the perception of space from a fundamental sensorimotor perspective. Despite its pervasive nature in our perception of the world, the origin of the concept of space remains largely mysterious. For example in the context of artificial perception, this issue is usually circumvented by having engineers pre-define the spatial structure of the problem the agent has to face. We here show that the structure of space can be autonomously discovered by a naive agent in the form of sensorimotor regularities, that correspond to so called compensable sensory experiences: these are experiences that can be generated either by the agent or its environment. By detecting such compensable experiences the agent can infer the topological and metric structure of the external space in which its body is moving. We propose a theoretical description of the nature of these regularities and illustrate the approach on a simulated robotic arm equipped with an eye-like sensor, and which interacts with an object. Finally we show how these regularities can be used to build an internal representation of the sensor's external spatial configuration.

A Formal Account of Structuring Motor Actions With Sensory Prediction for a Naive Agent

“…As a first step, this subsection is devoted to the introduction of all the notions and definitions of the motor side of the proposed sensorimotor framework. After highlighting the limitations of the previous approaches, as seen in e.g., Marcel et al (2017) and Laflaquière et al (2018), we show how to reparameterize the sensorimotor interaction by introducing motor actions. Their definition and properties are then carefully discussed.…”

“…This article follows much of the same approach begun in Philipona et al (2003) and subsequently developed in e.g., Marcel et al (2017), Marcel et al (2019), Laflaquière et al (2018), and Laflaquière and Ortiz (2019). In particular, it sets out to mathematically describe from an exterior, "objective, " point of view some properties of the interaction between the robot and its environment which should appear in its sensorimotor flow.…”

“…We already proposed different contributions in this field, successively dealing with peripersonal space characterization (Laflaquière et al, 2015), self-contact and body representation (Marcel et al, 2017) and the emergence of a topological representation of sensors poses (Marcel et al, 2019). These works, as well as Laflaquière et al (2018) and Laflaquière and Ortiz (2019), devote a significant effort to providing formalisms suited to make explicit, and, where applicable, formally prove-not only the processing required to capture the contingencies, i.e., invariants in the sensorimotor flow of the agent, but also the mechanisms by which said contingencies should appear. This is an attempt to pinpoint the exact conditions of validity of the proposed processes in order to deliver on the promises of genericity of SMCT.…”

The Language of Vision*

“…These regularities would also generate a grammar, specifying, among other rules, for example, that an action comes with an agent. Similar proposals of an emergence of structure from regularities have been proposed for other aspects of vision such as learning the metric structure of the world through regularities in photoreceptor responses as an object moves across them (Maloney & Ahumada, 1989) or how interactions with the world change the sensory input (Laflaquière et al., 2018; Philipona et al., 2003). In the proposal here, rather than just extracting the structure of space, language acquisition would have to extract the conceptual structure of the dynamic, visual world.…”

Space Emerges from What We Know—Spatial Categorisations Induced by Information Constraints

“…When one takes an even closer look at this and considers space strictly through the perspective of the sensorimotor loop as in [ 8 , 9 , 10 , 11 ], whatever spatial modalities exist, exist exclusively through what can be sensed and acted upon throughout the closed perception-action loop (see also [ 12 ]). The “space” emerging from this interaction is purely defined in terms of what is done vs. what is sensed.…”