Sensor Fusion and Information Sharing for Automated Vehicles in Intersections

Abstract: One of the biggest challenges in the development ofautonomous vehicles is to anticipate the behavior of other roadusers. Autonomous vehicles rely on data obtained by on-boardsensors and make decisions accordingly, but this becomes difficultif the sensors are occluded or have limited range. In this reportwe propose an algorithm for connected vehicles in an intersectionto fuse and share sensor data and gain a better estimationof the surrounding environment. The method used for sensorfusion was a Kalman filter and a tracking algorithm, where timedelay from external sensors was considered. Parameters for theKalman filter were decided through measurement of the sensors’variances as well as tuning. It was concluded that the variancesare dependent on the objects’ movements, which means thatconstant parameters for the Kalman filter would not be enoughto make it efficient. However, the tracking and the sensor sharingmade a significant difference in the vehicle’s detection rate whichcould ultimately increase safety in intersections.