360 Multisensor Object Fusion and Sensor-based Erroneous Data Management for Autonomous Vehicles

摘要

In the field of autonomous vehicle, in order to raise the road safety, multiple sensor technologies are mounted with varying levels of maturity. Multisensor data fusion is the process of combining observations from different sources to provide a robust and complete description of an environment and to overcome the limitation in term of availability (compared with one sensor), robustness and quality. Management of erroneous and incomplete information is an important requirement for perception systems. A robust and scalable 360° multisensor fusion framework for static and dynamic obstacles in conjunction with a sensor-based erroneous management block is proposed. The real time quality-based process considers the intra-sensor coherence and source antecedents to deal with the false positive in order to avert unwanted breaking or undesirable longitudinal control behaviours. Our framework is directly integrated on autonomous VALEO and OEM Customers democars. The evaluation is made using real data from different driving scenarios and proved its efficiency and robustness.