John Pisokas Title: ------ "Subsymbolic Action Planning for Mobile Robots" Abstract: --------- "The ability to determine a sequence of actions in order to reach a particular goal is of utmost importance to mobile robots. One possibility is to follow the symbolic paradigm; to identify the fundamental elements of behavioural space (the actions) manually, to specify the permissible operators applicable to them manually, and to use standard search techniques to find paths through the search space that will lead from an initial state to a goal state, as demonstrated for instance in STRIPS. Such an approach has several shortcomings: -1- Because of the external identification of operators, only part of the behavioural and perceptual space available to the robot may be exploited. -2- Actual robot perceptions and actions may differ from those defined by the human designer, eventually leading to brittle and unreliable performance. An alternative would be to let the robot acquire its own subsymbolic world representation, and to let it associate preconditions and postconditions with its behavioural repertoire autonomously. This presentation addresses the question of how a mobile robot can autonomously determine task-achieving sequences of actions, not relying on (externally supplied) symbolic representation, but on an acquired subsymbolic representation of perceptual and behavioural space, as the robot experiences it."