Planning of Autonomous and Mobile Robots in Dynamic Environments

The capabilities of autonomous robots are increasing which is a reason for research progress in the last years. An increasing number of robotic scenarios are becoming more viable in various domains for example, mowing, harvesting, warehouse robots. But autonomous mobile robots need more than the ability to navigate through their world, and manipulate objects. Even in teams of robots, a structured plan for each team member is necessary. Planning becomes a more important task. Furthermore, the environment can change, which results in failures for a robotic team using static plans. The team can end insolvable situation even when a robot breaks down. Automated planning helps to avoid failing behaviour caused by environment changes or robotic failures. We present a solution for planning heterogenous, autonomous, mobile robots in soft realtime, the proposed framework, pRoPhet MAS, based on the modeling language ALICA. This language enables description of the team activities from a central point of view. These team activities are realised by state machines annotated with tasks. ALICA supports basic algorithms for role and task assignment. The realisation of planning requires description of the logic. Hence, ALICA is extended by a first-order logic description. pRoPhEt MAS introduces a planning problem element in ALICA based on the Planning Problem Definition Language (PDDL). Planning problems can be solved offline to support the behaviour modelling, online to work in reactive domains, or interactively as guidance in extraterrestrial scenarios. The planner generates a multi-agent plan based on a goal description in the planning problem element where these plans require a specific task and role allocation algorithm. The online planning uses a soft real-time limit when providing promising plans. If a created plan fails while executing, pRoPhet MAS starts to replan. Finally, the pRoPhet MAS can divide the planning process among the teammates. We evaluate our approach in the research project IMPERA (Integrated Mission Planning for Distributed Robot Systems) as proof of concept. The usage of the planner is demonstrated as mission guidance for a team of four robots consisting of three different types. Moreover, pRoPhEt MAS successfully controls five robots in a reactive blocks world domain wherein the team was disturbed by opponent robots to produce plan failures. Then, we evaluate the parallel planning engine MAGiC for a complex planning problem using low communication overhead. Finally, we evaluate the task- and role-allocation algorithm towards increasing agents and tasks.