A generic multi-lidar data batching strategy on the sensor driver level

Abstract. This paper addresses how to utilize multiple spinning lidar sensors for real-time applications. Especially how to derive back the problem to having only a single lidar input, to which there are countless available algorithms solving odometry, mapping, object detection and tracking and many other tasks. We provide a strategy that can be implemented to most if not all spinning lidars on the market. Instead of traditional data batching that accumulates data packets based on the spinning angle, we propose batching based on the sampling time, which also enable us to ensure strict time alignment within the multiple lidar sources. In order to demonstrate our batching strategy, we provide a case study where we evaluated a SLAM algorithm with a single and a dual-lidar setup. Our batching algorithm enabled us to use the SLAM algorithm that was previously designed for a single spinning lidar without any additional change, while it showcased benefits, especially in stability due to the larger field of view and reduced occlusion.