Offload scale optimization aided SLAM using reinforcement learning for digital twin

The level of safety required for autonomous driving has increased as autonomous driving technology has been developed. A safety integrity level of this magnitude necessitates adequate precision and accuracy against environmental recognition such as using artificial intelligence(AI). The requirements for the amount of computation cost have been raised alongside the development of artificial intelligence or core functionalities regarding autonomous driving. As the amount of computation cost increases, it becomes difficult to perform computation tasks with a limited performance of the computer on the vehicle. In order to solve this limited computation issues, cloud-based offloading is considered a potential candidate for a feasible solution. However, when transmitting datasets to the cloud through a network, there is a problem that the scale of datasets exceeding the network bandwidth causes transmission latency. This study introduces a novel cloud offloading approach that uses scale optimization in master-slave architecture based on a real-time simultaneous localization and mapping case scenario. The result shows that mitigation of communication latency issues through scale of datasets optimization. It is expected that the suggested framework can be aided in the cost reduction of implementing multiple agents and vehicular applications, especially using a cloud-based approach.