Objects Perceptibility Prediction Model Based on Machine Learning for V2I Communication Load Reduction.

Autonomous driving is becoming prevalent and important for enhancing road safety. Sensor-based perception on vehicles is the main technology in autonomous driving systems. The environmental perception by individual vehicle has the limitations on coverage and detection accuracy. This issue can be solved by sending environmental information from the roadside infrastructure to autonomous vehicles. However, it results in a heavy communication load. In this paper, we present a machine learning based model to predict whether an autonomous vehicle can perceive an object. The prediction result can help roadside infrastructures to send the cooperative information selectively, and finally reduces the network load. The experiments prove that the accuracy of the model reaches 93%, and the V2I communication load reduction is up to 55%.