Environment-aware Optimization of Track-to-Track Fusion for Collective Perception

Correct and complete perception is key for autonomous vehicles to plan safe maneuvers. Especially under harsh weather conditions the use of sensing capabilities from other road users via vehicle-to-everything (V2X) communication can contribute to more complete perception. However, information from other road users may contain additional uncertainties and lead to less accurate perception. Additionally, attackers may use the V2X channel to transmit malicious data. For building an accurate environmental model an autonomous vehicle needs as precise information as possible. To tackle the problems of additional uncertainties within collective perception we propose a methodology to check perceived information for its trustworthiness and validity. This is achieved by evaluating the perception capabilities of a holistic perception pipeline and checking collectively transmitted information for consistency. The proposed approach is evaluated under varying environmental conditions on a simulated highway scenario.