Robot task planning based on resource reasoning

A planning paradigm based on resource reasoning is presented. The notion of resource is extended to include not only materials and machines but also space and time, and planning is considered as the management of resource distribution to achieve a goal. Each operator in PLARE (a planner with resource reasoning) is considered a device which consumes a resource as input and produces a resource as an output. PLARE keeps track of the change of each resource with its own resources window while generating a plan. Whenever action operators are competing for a resource/resource and there ar demands than supplies, a resource conflict is detected. Since it is too complicated to consider all the resource conflicts in the middle planning, a least-commitment strategy is adopted in dealing with resource conflicts: PLARE makes a plan in the bidirectional way; backward planning for singular resources is followed by forward planning for plural resources. After a final backward plan is generated for singular resources, PLARE starts distributing plural resources forwardly to the plan network