Collecting Simulation Scenarios by Analyzing Physical Test Drives
Abstract
Ensuring safe operation of autonomous vehicles requires testing them including critical combinations of obstacle configurations plus sensor and actuator inaccuracies. A method for testing inaccuracy combinations has already... [ view full abstract ]
Ensuring safe operation of autonomous vehicles requires testing them including critical combinations of obstacle configurations plus sensor and actuator inaccuracies. A method for testing inaccuracy combinations has already been published by the authors. This paper enhances the capabilities of the method by automatically collecting scenarios from physical vehicle drives that are relevant for further analysis. For such situations, a state trace including all variables of the whole planning and control system is stored together with environment information. The stored data is the input for further analysis. An implementation of this approach is demonstrated using a simulation and a full size vehicle.
Authors
-
Pascal Minnerup
(fortiss GmbH)
-
Tobias Kessler
(fortiss GmbH)
-
Alois Knoll
(Technische Universität München)
Topic Areas
Data Mining and Data Analysis , Driver Assistance Systems , ITS Field Tests and Implementation , Sensing, Vision, and Perception , Simulation and Modeling
Session
Fr-B8 » Data Mining and Data Analysis VII (13:40 - Friday, 18th September, San Borondón B2)