Control of robots, and especially mobile robots, is one of the traditional testbeds for problem solving, planning, and other areas of artifical intelligence. As long ago as the 1960s, SRI's Shakey was used to develop and refine artificial intelligence algorithms [10]. A survey of mobile robots through the 1980s can be found in[14].; some more recent examples are in [13].
Although originally intended as a model of human problem-solving processes, rule-based systems have been used successfully in implementing reactive control strategies (ie , control strategies in which the robot exhibits behavior as reactions to events [20]) for mobile robots. The use of a rule-based system to implement a reactive strategy has been proposed in, for instance, programming NASA's Mars rover [9]. An alternative approach to implementing a reactive strategy for planetary explorers is in [6]. General information on reactive real-time systems can be found in [15].
A particular form of reactive robot control using a hierarchy of rules
is Brooks's subsumption architecture
[5].
In a
subsumption architecture, rules are assigned a priority with
high-priority rules capable of overriding low-priority rules that
would otherwise be enabled. This allows the programmer to develop
several levels of competence for the robot; the lower priority rules,
associated with lower levels of competence, define behaviors such as
line following, while the higher priority rules, associated with
higher levels of competence, define more complex behaviors such as
mapping or the solution of subgoals. As such, higher level behaviors
subsume lower level behaviors.