Collective Perception – Vehicular Communication and Data Fusion

Autonomous driving is mostly supported by onboard sensors. Vehicle-to-everything (V2X) allows close road users to communicate with each other in real-time. V2X enables vehicles to work together as they are constantly "talking" with each other. Road-Side Units (RSU) could also share information with surrounding vehicles. Some V2X services like Cooperative Awareness (CA) and Decentralized Environmental Notification (DEN) have already been standardized, while the Collective Perception Message (CPM) is still under progress. By sharing raw or processed sensor data with neighbor vehicles, those can perceive objects beyond their sensors‘ detection range and increase the sensing accuracy. The bandwidth dedicated to CPM is limited though, so that even with processed data, it may get saturated when more vehicles send simultaneously. A Decentralized Congestion Control (DCC) mechanism should mitigate or avoid packet collision. When the CPM is correctly received, its data needs to be integrated with the data from onboard sensors using a data fusion.

Researchers have proposed various methods to increase the Packet Delivery Ratio (PDR), using DCC concepts and CPM generation rules. However, the methods have not been validated for both communication and perception together. Another important aspect is to ensure fairness among the exchanging vehicles, so that all are able to send at some time.

In this doctoral project, CPM generation rules and DCC will be validated regarding communication and perception in a common simulator setup with RSU and V2X vehicles. Carla vehicle simulation is used for realistic perception and is combined with Artery to generate CPM, while their transmission is simulated through OMNET++. SUMO generates vehicle traffic and synchronizes it with the other components.

Researcher: Shule Li, M. Sc.