Perception for action in roving robots: A dynamical system approach

Abstract

For a mobile robot working in the real world, the ability to interpret information coming from the environment is crucial, both for its survival and for its accomplishing tasks. The cognitive capabilities of a robotic system are defined by the way in which the information gathered from sensors are processed to produce a specific action or behavior. Two broad classes can be distinguished: the cognitivist approach based on symbolic information processing and the emergent systems approach that is directed to the application of dynamical systems connected to the principles of self-organization [39]. Among the numerous solutions proposed by researchers, a great part is located in between the two main approaches. In the recent past, research was directed towards endowing a robot with capabilities of self-creating an internal representation of the environment. Experiments in real world differ from applications in structured environments because they are mostly dynamically changing, so that it is impossible to program robot behaviors only on the basis of a priori knowledge. Moreover, the control loop must be able to process the different stimuli coming from the environment in a time that must be compatible with the real time applications. To solve these open problems, research activities have been focused on suitable solutions obtained taking inspiration from nature and applying biological principles to develop new control systems for perception-action purposes. © 2009 Springer London.