From abstract to executable plans

Autor: Wiesner, Robert
Přispěvatelé: Barták, Roman, Švancara, Jiří
Jazyk: čeština
Rok vydání: 2020
Předmět:
Popis: This thesis delves into postprocesing on plans for Multi Agent Path Finding. The intent behind postprocesing is to chang the plans so they can be performed by real robots. Changes include reduction in wasted time and introduction of robustness. This thesis proposes five algorithms (later extended to ten), through which MAPF plans are being transfered to a significantly lower level of abstraction. Results were tested on sixty plans created by MAPF solver MAPF Scenatio over five maps. Some resulting plans were tested in real life on robots Ozobot Evo. Results indicate that significantly lower makespan and greater robustness can be achieved through postprocessing. This study also indicates that various algorithms are more effective for certain types of plans.
Databáze: OpenAIRE