Behavior-based obstacle detection in off-road environments considering data quality

Abstract

In off-road environments, the assessment and classification in travers-able and non-traversable areas is a challenging task. Not only does the drive-ability depend on the vehicle’s state in combination with the environmental geometry, but also the assessment is complicated by noisy and wrong sensor data. Thereby, faulty evaluation may lead to severe harm to goods or people since either safety issues or reliability problems are caused. While for the control part behavior-based systems proved to be suitable due to their inherent robustness against unforeseen situations, robust perception is still an unsolved problem leading to severe system failures. This paper faces the perception problem by introducing a new data quality-based perception module based on the integrated Behavior-Based Control (iB2C) architecture. Therefore, a new concept of data quality in behavior-based systems and methods for quality aware data fusion are developed while taking advantage of the modularity, extensibility and traceability of the existing architecture.