Introduction

Abstract

Artificial spatial agents, such as the mobile robots we will be concerned with in this book, have a great deal in common with their natural counterparts, animals and humans. They are all embodied physical systems situated in the real world, important characteristics regarded by many as a prerequisite for genuine intelligence (Pfeifer & Bongard, 2007; Varela et al., 1992). They share the ability to perceive their environment and extract spatial information from their perceptions. They store spatial information over time and this information affects their future decisions and actions. And they are able to affect the state of their environment by their actions. An important part of acting in space involves moving to other parts of the environment outside of the agent’s current sensory scope and navigation between known places. Hence, a spatial agent benefits from the ability to integrate local observations and to derive spatial relationships on a larger scale.

The details of how spatial information is extracted, stored, and processed in humans and animals are still largely unclear and subject of ongoing research across many disciplines. In this book, we follow the tradition of Braitenberg’s “law of uphill analysis and downhill invention” (Braitenberg, 1984) in the sense that we attempt to design artificial agents that demonstrate a certain set of competences, possibly drawing inspiration from empirical studies about how these competences are achieved by humans or animals. The results, positive and negative ones, can then be used to draw conclusions about spatial information processing in natural agents.