Autonomous HTN planning algorithm with upward-backtracking mechanism infused

Hierarchical task network (HTN) is widely used for intelligent planning. For typical HTN algorithm, if the current state does not support any decomposition prescription, the HTN planner will not try to change the state, but just return failure. This characteristic requires technicians to compile a large prescription database, which is time consuming. To overcome this shortcoming, the paper infuses upward-backtracking mechanism to improve HTN algorithm. When the current state does not support any decomposition method, the improved planner will search higher level tasks and its subtasks to change the system state to support the decomposition of the mission. The promotion will extend the searching range of planner and reduce the compiling of prescription database largely. A task of structure machining is taken as an example to demonstrate the algorithm´s property.