Incentivizing Smart Vehicles For Autonomous Driving HD Maps using predictive analytics

Recent advancements in distributed data storage, colloborative processing capabilities, and their architectural design have a profound influence on our daily lives, simplifying the process of hosting computation-intensive software applications. A single software hosting device, for instance, an Onboard Computation Unit (OBCU) of an autonomous vehicle, on the other hand, can’t keep up with the growing need for diverse requirements of computing architectures and processing power of such applications. For example, the real-time handling and generation of HD maps is one such crucial and highly computation-intensive application for autonomous vehicles. For this purpose, an efficient solution is to rely on the computing power of adjacent resource-rich nodes. Hence, the use of a latency-aware predictive analytics-based computation task offloading method has a vital significance, especially with the advent of Vehicular Edge and Fog Computing (VEFC). VEFC models pool redundant computational resources in close proximity. It, however, faces several challenges. The stochastic nature of vehicular networks, intricate heterogeneity at several levels, and the uncertainty of time-sensitive applications for the timely completion of tasks impede the VEFC’s efficiency. We developed a computation offloading framework that timely alleviates the computing-resource-deficit devices by deploying vehicles as fog-nodes commensurate with the minimum latency requirements of advanced vehicular applications, especially computation-intensive HD map handling, and generation in real-time. Our technique is a greedy heuristic method that works in small, discrete steps instead of solving the entire optimization problem holistically. We used Monte Carlo simulation to show that our approach had an overall response time of under-300 milliseconds compared to other baseline methods.