Merits and Limitations of Automotive Radar for Land Vehicle Positioning in Challenging Environments

Land vehicles of the near future require accurate positioning systems that are robust across diverse and changing environmental conditions. While global navigation satellite systems (GNSSs) remain standard for absolute positioning on land, access to satellite signals is unreliable or absent in many urban environments. Inertial navigation systems (INSs) can provide positioning solutions capable of bridging short GNSS outages but cannot sustain adequate positioning accuracy for the duration of outages often present in cities. The need for vehicles to navigate reliably through such environments has motivated research into multisensor fusion for positioning. Aiding sensors include cameras, light detection and ranging (LiDAR), and radar. Due to the varying effects of environmental conditions on each type of sensor, more than one sensor system must be implemented. Radars are an attractive automotive sensor due to their insensitivity to adverse lighting conditions, which affects cameras, and inclement weather, which impacts both cameras and LiDAR. Automotive radars are low-cost sensors found in most modern vehicles, applied widely for driver assistance systems. Recent advancements in radar technology, however, have led to research in radar-based positioning. This article presents a comparative case study and analysis of two radar-based positioning methods across three practical driving scenarios. Radar odometry and radar-to-map registration are applied to real driving scenarios, including a university campus, a busy shopping street, and an indoor parking garage. Areas where radar-based positioning fails are discussed, aiming to identify challenges faced specifically by automotive radar. In addition, areas where radar-based navigation is already performing robustly are presented.