Automatic Construction of Reactive Plans for A Box-Pushing Task of A Mobile Robot

모바일 로봇의 상자 밀어내기 작업을 위한 리엑티브 플랜의 자동 생성

Reactive plans for a box-pushing task of a mobile robot are automatically built up, where sequential action plans are found in a configuration space by A* algorithm for various initial configurations. Then, conjunction of conditions to associate with a same behavior are found by a back tracking algorithm. And corresponding reactive plans are generated. Finally, a clustering technique is applied to identify which reactive plan should be applied for a given perceptual condition. Several simulation results are shown to justify our proposed approach.