17 January 2008
A multi-threaded architecture for cognitive robotics
12:30pm - 13:45pm
Prof Keith Clark - Imperial College
We describe a multi-threaded architecture combining elements of Belief, Desires, Intentions (BDI) agent architectures with the goal directed reactive control of Nilsson's Teleo Reactive (TR) programs.
TR programs are a rule based programming notation influenced
by process control concepts of continuous monitoring and action, but they have parameterized procedures and actions can be TR procedure calls, even recursive calls. A TR procedure is also typically goal directed, the actions of its rules being oriented towards achieving a state of the environment that can be determined by some test of sensor readings. They are thus well suited to implementing robotic control where action routines are selected on the basis of the computational analysis of sensor readings .
TR computation comprises nested threads of execution, each thread being the execution of a TR procedure called from its parent thread. The TR procedures of the higher threads can be programmed to monitor beliefs, inferred from sensor readings and a model of the environment, rather than percepts. But to retain reactivity, the lower level threads can still directly test percepts computed from sensor readings. This makes it also suitable for programming hybrid deliberative/reactive robots. Finally, a topmost control thread, which decides which high level TR procedure to invoke based on significant events, such as belief updates or new goals, can be added borrowing from BDI agent architectures. This allows the robot to switch tasks based on message events or sensor reading events that leading to significant changes in the robots higher level beliefs. This is the control architecture currently being used at Imperial both for Multi-agent systems and co-operative robotics applications.
Professor Keith Clark studied Mathematics, then Philosophy, before switching to Computing for his PhD. He joined the Department of Computing, Imperial College, in 1979 and was appointed Professor of Computational Logic in 1988. He is a visiting Professor at the University of Queensland in Australia, and Blekinge Institue of Technology in Sweden. In 1980 he co-founded Logic Programming Associates to develop and market Prolog systems for personal computers.
Prior to 1980 his research was in the are of computation logic. His research in the 1980s was in the area of logic programming languages, their concurrent and distributed implementation, and their applications. Since 1990 his interests have focussed more on multi-threaded symbolic programming languages, still with a declarative emphasis, and their use for building multi-agent and cognitive robotics applications. Joint recent research has covered: a generic multi-threaded BDI agent/robot control architecture, personal recommendation agents, agents for automatically extraction of information from the web and its display via graphical user interfaces, norm constraints for agents interacting within a institutional environment and agents partly programmed by a set of formal bi-lateral contracts.
Save to your Calendar