Where Should We Place LiDARs on the Autonomous Vehicle? - An Optimal Design Approach
Considering its reliability to provide accurate 3D views along with precise distance measures under highly uncertain driving conditions, automakers deems LiDAR a critical sensor for autonomous vehicle (AV) applications. However, its monumental cost restricts the practical implementations and surges many AV designers to best calibrate the sensor configurations to fully utilize its capability. This paper investigates the optimal LiDAR configuration problem to achieve such utility maximization purposes. The perception area and non-detectable subspace are used to construct the min-max optimization problem and a bio-inspired measure -- volume to surface area ratio (VSR) -- is proposed as an easy-to-evaluate cost function representing the notion of the size of the non-detectable subspaces of a given configuration. A cuboid-based approach is then adopted to show that VSR-based measure is a well-suited proxy for object detection rate and to yield tractable cost function computation using artificial bee colony evolutionary algorithm. Our experiments highlight the effectiveness of this approach to prescribe the cost-effective configuration of the LiDARs and to provide insightful analyses for automakers and suppliers to better design AV systems.
READ FULL TEXT