We are developing a perception system for truck platooning, where a radar and a camera sensors are primarily used for the perception task[1]. The perception task in truck platooning is challenging due to its unique driving condition with trucks driving at high speeds maintaining very close gaps between trucks. Because of the close distance, the available sensor information is much too limited compared with normal driving scenarios. To overcome this hurdle, we use (i) deep learning-based vehicle detection and lane detection, (ii) sensor fusion with a radar sensor. More specifically, a dynamic region of interest (RoI) is calculated from the information from the camera and radar sensors to precisely filter out noisy input data from the occlusion effect by the front truck.

Also, we are developing a 1/14 scale truck platooning testbed[2] that can be used to test various platooning operations such as joining of a new truck and merging and splitting of multiple truck platoons. Besides, our testbed can be used for testing various safety measures such as gradual lane change of multiple trucks with surrounding vehicles in an emergency scenario.