A STUDY FOR THE APPLICATION OF AUTOMATED PLANNING TO MOBILE ASSISTIVE ROBOTS

In recent years, Automated Planning (AP) has experienced important advances. In this study we apply such advances to the field of Mobile Assistive Robots (MAR). In particular, we propose the use of AP to implement the deliberative step between observation and action execution in MAR. First, we analyze the requirements that allow a MAR to plan navigation and manipulation actions in near real time. The intention is to build the foundation for a planning module within the Simultaneous User Learning and TAsk executioN (SULTAN) architecture, allowing a MAR to perform Daily Life Activities (DLA) in humanlike environments. Second, we apply AP techniques in fully observable, deterministic and static simulated environments with a single MAR. In addition, we analyze and compare the best available satisficing automated planners. The selected planners participate in several experiments to obtain plans for a Planning Domain Definition Language (PDDL) based on the Tidybot domain. Finally, in order to know how competitive the selected planners are, we compare the experimental results in detail.