Distributed Environmental Information Management System for Autonomous Vehicles using Edge Computing
Shunya Yamada, Yousuke Watanabe, Hiroaki Takada · Journal of the Robotics Society of Japan · 2020
In order for autonomous vehicles to drive safely and comfortably, environmental information detected by sensors need to be gathered from wider area and to be more accurate. We can improve data accuracy by fusions of sensor data from not only vehicles but also road infrastructures. Fusion processing is usually performed in a high-performance server (centralized system). However, when the number of sensors is enormous, processing time and communication time for fusions become unacceptable due to high-load and limited capacity of network. And, waiting data-arrivals from all sensors is impractical in such situations. Thus, fusions should be distributed and incrementally updated on each data-arrival. In this paper, we propose a distributed environmental information management system using edge computing and a sensor fusion method.Since the system is composed of geographically distributed edges and a centralized cloud, it can distribute processing costs and communication costs of fusions to the edges and the cloud. The proposal sensor fusion method can incrementally compute intermediate results without waiting for receiving all the environmental information. In comparison experiments with the centralized system, the proposed system improved the efficiency of data processing and reduced the amount of communication data.