Task planning in an uncertain world

Given that so much is known about the issues involved in classical task planning and given the requirement for some form of predictive planning in a semi-autonomous or autonomous robot, the use of standard techniques is highly desirable. However, classical planners make strong and inappropriately restrictive assumptions about the world on which the planner operates which must be modified if a planner is to be usefully applied in robotic domains. The authors describe the work done in developing a prototype task planner called HELP (Hierarchical Execution-Led Planner) for such domains, and the way in which the classical approach has been adapted to cope with the relaxation of these assumptions.<>