In this paper we propose and justify a methodology for the development of the control systems, or `cognitive architectures', of autonomous mobile robots. We argue that the design by hand of such control systems becomes prohibitively difficult as complexity increases. We discuss an alternative approach, involving artificial evolution, where the basic building blocks for cognitive architectures are adaptive noise-tolerant dynamical neural networks, rather than programs. These networks may be recurrent, and should operate in real time. Evolution should be incremental, using an extended and modified version of genetic algorithms. We nally propose that, sooner rather than later, visual processing will be required in order for robots to engage in non-trivial navigation behaviours. Time constraints suggest that initial architecture evaluations should be largely done in simulation. The pitfalls of simulations compared with reality are discussed, together with the importance of incorporating noise. To support our claims and proposals, we present results from some preliminary experiments where robots which roam office-like environments are evolved.

A preliminary unedited version of this paper was incorrectly published as part of Volume

We present an architecture for robot control which can be viewed as a very fine-grained layered architecture motivated by the principles underlying the subsumption architecture. The subsumption architecture provides a powerful means for defining intelligent robot control mechanisms through the layered composition of simple behaviors. However, we have found that there are basic limitations inherent in this architecture due to the inaccessibility of information internal to a behavior. While adhering to the basic concept of building a robot control system through successive layers of competence, task achieving behaviors in our system are fragmented into many smaller decision-making units. Each of these units simply has the task of transforming a set of input activations into an output activation, so that the role that any unit plays in the system is defined entirely by how it is connected to other units. This fine-grained nature of our architecture permits a more flexible arbitration of c...

We present an expressive agent design language for reinforcement learning that allows the user to constrain the policies considered by the learning process.The language includes standard features such as parameterized subroutines, temporary interrupts, aborts, and memory variables, but also allows for unspecified choices in the agent program. For learning that which isn’t specified, we present provably convergent learning algorithms. We demonstrate by example that agent programs written in the language are concise as well as modular. This facilitates state abstraction and the transferability of learned skills. 1

This technical definition will only make sense toe reader by Ch. 4, once goals and management processes have been described. All that matters forrs section is that a difference between goals and perturbance be noted by the reader. Astate perturbance is not a goal, but it arises out of the processing of goals. In Ch. 7, arelation00 perturbance and "emotion" is discussed. 43 . Sloman says of certain moods that they are "persistent states with dispositional power to color and modify a host of other states and processes. Such moodscan39061-6 be caused by cognitive events with semantic content, though they need not be.[...]0-64000 their control function does not require specific semantic content, though theycan0371-62 cognitive processes that do involve semantic content." (Sloman, 1992b Section 6).A 39642 view is taken in (Oatley, 1992). To be more precise, moods are temporary control stateswhich9881-5 the prominence of some motivators while decreasing others. In particular, they affectthe 41330-5 that certain "goal generators" are triggered. Moreover, moods affect the valenceofce 39476 evaluations, and the likelihood of affective evaluations (perhaps by modifying thresholdsofsholds 42 that trigger evaluations). It is not yet clear whether moods as defined here are9531 - or whether they merely emerge as side-effects of functional processes. . A reflex is a ballistic form of behaviour that can be specified by a narrow setw rules based on input integration and a narrow amount of internal state. There aretwo0981 of reflexes: simple reflexes and fixed action patterns. A simple reflex involves oneaction,-43000 a fixed action pattern involves a collection of actions. Usually, at most only asmall-4120 of perceptual feedback influences reflex action. This would require a definit...

This paper is concerned with approaches to computational modelling of the neural mechanisms underlying behaviour. It examines the relationship between computational neuroscience (e.g. [52, 34]) and that style of modelling popularly referred to as "connectionism", "parallel distributed processing" (pdp), or "neural networks"

Active cameras provide a navigating vehicle with the ability to fixate and track features over extended periods of time, and wide fields of view. While it is relatively straightforward to apply fixating vision to tactical, short-term navigation tasks, using serial fixation on a succession of features to provide global information for strategic navigation is more involved. However, active vision is seemingly well-suited to this task: the ability to measure features over such a wide range means that the same ones can be used as a robot makes a wide range of movements. This has advantages for map-building and localisation. The core work of this thesis concerns simultaneous localisation and map-building for a robot with a stereo active head, operating in an unknown environment and using point features in the world as visual landmarks. Importance has been attached to producing maps which are useful for extended periods of navigation. Many map-building methods fail on extended runs because ...

Both direct, and evolved, behavior-based approaches to mobile robots have yielded a numberofinteresting demonstrations of robots that navigate, map, plan and operate in the real world. The work can best be described as attempts to emulate insect level locomotion and navigation, with very little work on behavior-based non-trivial manipulation of the world. There have been some behavior-based attempts at exploring social interactions, but these too have been modeled after the sorts of social interactions we see in insects. But thinking how to scale from all this insect level work to full human level intelligence and social interactions leads to a synthesis that is very di erent from that imagined in traditional Arti cial Intelligence and Cognitive Science. We report on work towards that goal.

Intelligent robots should be able to use sensor information to learn how to behave in a changing environment. As environmental complexity grows, the learning task becomes more and more difficult. We face this problem using an architecture based on learning classifier systems and on the structural properties of animal behavioural organization, as proposed by ethologists. After a description of the learning technique used and of the organizational structure proposed, we present experiments that show how behaviour acquisition can be achieved. Our simulated robot learns to follow a light and to avoid hot dangerous objects. While these two simple behavioural patterns are independently learnt, coordination is attained by means of a learning coordination mechanism. Again this capacity is demonstrated by performing a number of experiments. 2 I. Introduction The traditional knowledge-based approach to artificial intelligence shows some fundamental deficiencies in the generation of powerful ...

This thesis addresses the problem of intelligent control of autonomous mobile robots, particularly under circumstances unforeseen by the designer. As the range of applications for autonomous robots widens and increasingly includes operation in unknown environments (exploration) and tasks which are not clearly specifiable a priori (maintenance work), this question is becoming more and more important. It is argued that in order to achieve such flexibility in unforeseen situations it is necessary to equip a mobile robot with the ability to autonomously acquire the necessary task achieving competences, through interaction with the world. Using mobile robots equipped with self-organising, behaviour-based controllers, experiments in the autonomous acquisition of motor competences and navigational skills were conducted to investigate the viability of this approach. A controller architecture is presented that allows extremely fast acquisition of motor competences such as obstacle avoidance, wa...