Hierarchical Organization of Robots

2 Robotic Organization

Many studies into robot group behaviour are done within the field of behaviour-oriented robotics and artificial life, focusing on how complex patterns of 'social behaviour' can emerge from local interaction rules in a group of homogenous robots. Such work is interesting in applications where robust collaborative behaviour is required and where specific skills or 'intelligence' of single robust is not required. The term 'collective behaviour' is used for such a distributed form of intelligence. Social insect societies (e.g. ants, bees, termites) have been used as biological models. (Deneubourg et al 1991) give an impressive example where a group of ant-like robots collectively 'solves' a sorting task. Their work is based on a model of how ants behave, using the principle of 'stigmergy' which is defined as "The production of a certain behaviour in agents as a consequence of the effects produced in the local environment by previous behaviour" (Beckers et al 1994). Mataric (1995) gives an overview on designing collective, autonomous (robotic) agents. Principles of collective behaviour are usually applied to a group of homogeneous robots which do not recognise or treat each other individually, i.e. they do not use any representations of other agents or explicit communication. In contrast, the term 'cooperation' describes a form of interaction which usually uses some form of more advanced communication. "Specifically, any cooperative behaviors that require negotiation between agents depend on directed communication in order to assign particular tasks" (Mataric 1995). Different 'roles' between agents are studied in (Kelly et al. 1997), a flocking behaviour where one robot is the leader, but the role of the 'leader' is only temporally assigned and depends on local information only. Moreover there is only one fairly simple 'task' (staying together) which does not change.

Behaviour based research on the principle of stigmergy is not using explicit representations of goals, the dynamics of group behaviour are emergent and self-organising. The results of such behaviour can be astonishing (e.g. building activities or feeding behaviour of social insects), but is different from highly complex forms of social organisation and cooperation which we find e.g. in mammal societies (see hunting behaviour of wolves or organisation of human society), employing division of labour, individual 'roles' and tasks allocated to specific individuals, and as such based on hierarchical organisation. Hierarchies in mammal societies can be either fairly rigid or flexible, adapted to specific needs and changing environmental conditions. Bases of an individualised society are particular relationships and explicit communication among individuals.

Hierarchical organisation of machines used in, for example, industrial applications are like master-slave approaches where a master computer (the planning and reasoning unit) is allocating tasks to a group of machines. In teleoperation approaches a master (human) - slave (robot) architecture is also widely used. Applied to groups of autonomous robots (Dautenhahn 1995) hierarchical organisation could be useful in a variety of applications, e.g. in a situation where a heterogeneous group of agents is collaborating on a common task. Let us assume a group of robots with special skills and behavioural capacities, e.g. a group of experts robots which has to function and 'survive' autonomously in a sewage pipe system or other dangerous/remote areas where human operators are not available. In such a group of heterogeneous robots it is very difficult to fix once in advance the hierarchy, i.e. in one situation robot A might be the leader, in another situation robot B would be most competent. Such situations, where (1) collaboration between agents is crucial for the functionality and survival of the group, and (2) tasks are complex and vary, in such scenarios we expect that an organizational structure which is revised periodically in light of results and experience can significantly enhance the adaptive value of the whole organisation, namely the group of autonomous robots.

Hierarchical Organization of Robots - 20 MAY 98

[Next] [Previous] [Top] [Contents]

Generated with CERN WebMaker