TY - GEN
T1 - A Fast Method for Explanations of Failures in Optimization-Based Robot Motion Planning
AU - Eder, Matthias Josef
AU - Steinbauer-Wagner, Gerald
PY - 2022
Y1 - 2022
N2 - The transparent interaction between an operator and a robot system is essential for successful task completion. This requires a mutual understanding of decisions and processes in order to provide accurate diagnoses and troubleshooting alternatives in the event of a failure. Due to inaccuracies in the environmental perception or planner configuration, errors can occur in robot motion planning that are hard to understand by the operator. In this work we present a method that is able to provide explanations for motion planning failures quickly. In the context of optimization-based planners, failures origin from planning constraints can be identified using an adaption of the FastDiag algorithm. It is able to provide one preferred minimal diagnosis in logarithmic time, also for large constraint sets. To evaluate the applicability of the proposed method, experiments are conducted that compare the computational performance to an existing method while considering different parameters such as number of constraints and requested diagnoses.
AB - The transparent interaction between an operator and a robot system is essential for successful task completion. This requires a mutual understanding of decisions and processes in order to provide accurate diagnoses and troubleshooting alternatives in the event of a failure. Due to inaccuracies in the environmental perception or planner configuration, errors can occur in robot motion planning that are hard to understand by the operator. In this work we present a method that is able to provide explanations for motion planning failures quickly. In the context of optimization-based planners, failures origin from planning constraints can be identified using an adaption of the FastDiag algorithm. It is able to provide one preferred minimal diagnosis in logarithmic time, also for large constraint sets. To evaluate the applicability of the proposed method, experiments are conducted that compare the computational performance to an existing method while considering different parameters such as number of constraints and requested diagnoses.
KW - Diagnosis
KW - Explainable artificial intelligence
KW - FastDiag
KW - Human-robot interaction
KW - Motion planning
KW - XAI
UR - http://www.scopus.com/inward/record.url?scp=85129291252&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-04870-8_14
DO - 10.1007/978-3-031-04870-8_14
M3 - Conference paper
SN - 9783031048692
T3 - Mechanisms and Machine Science
SP - 114
EP - 121
BT - Advances in Service and Industrial Robotics - RAAD 2022
A2 - Müller, Andreas
A2 - Brandstötter, Mathias
PB - Springer
T2 - 31st International Conference on Robotics in Alpe-Adria-Danube Region
Y2 - 8 June 2022 through 10 June 2022
ER -