An Incentivized Online Learning Approach for Context Information Caching and Sharing in V2X Networks

Provisioning context information via vehicle-to-everything (V2X) networks can greatly enhance the context awareness and driving intelligence of vehicles. Considering the location-related interests, it is favorable to employ the cache-enabled vehicles for content sharing. In this work, we investigate how to incentivize the selfish vehicles to share their context information with the multi-dimensional imperfect information of content dynamics, requests, and cache costs. The incentivized interaction process between the Base Station (BS) and cache-enabled vehicles is modeled as an online learning problem from the BS aspect. Based on the reverse auction theorem, an incentivized online vehicle caching mechanism is proposed to ensure the vehicles' voluntary participation in edge services and maximize the social welfare (vehicle-offloaded traffic excluding the caching and sharing cost), which is proved to approach an idealistic performance with prior knowledge of reward of contents and cost of vehicle caching. Simulation results show that the proposed incentive method can enhance the social welfare by around 3X to 7X compared with the popularity based and random caching schemes.