Sensor scan timing compensation in environment models for automated road vehicles
Jens Rieken, Markus Maurer · 2016
Environment perception is one of the most important tasks for Advanced Driver Assistance Systems and Automated Driving. Various types of sensors are used concurrently and fused into a consistent representation of the vehicle's surrounding. With more than one device involved, timing behavior becomes a relevant aspect of such systems. Delays due to processing and transferring data need to be considered when fusing sensor data. While some sensors, such as camera devices, perceive the environment as a snapshot, other sensors are based on a scanning technique. The iterative acquisition principle of scanning sensors lead to considerable time differences between single measurements within a scan. In addition to transfer and processing delays, these time differences have to be considered properly for a comprehensive environment representation. This paper focuses on time-related effects resulting from using a scanning sensor. The influence on two commonly utilized environment representations is discussed, grid-based and object-based approaches for stationary and movable elements, respectively. We present a comprehensive approach to incorporate the ego motion into grid-based models. We propose an adaptive prediction horizon for object tracking algorithms, based on sensor scan timing characteristics. Performance gains are evaluated with simulated data and verified within a real-world application. The presented algorithms are computationally feasible and real-time capable on a standard PC.