Short-Term Wireless Connectivity Prediction for Connected Agricultural Vehicles

Abstract

Robots and autonomous vehicles have been integrated in our life and utilized in a plethora of application scenarios, including intelligent transportation, industrial automation and smart agriculture. Several of the these applications might be functioning in environments where cellular network coverage is low or non-existent. In a case like this, lower bandwidth networks and vehicle-to-vehicle communication can be used to keep the application operating safely, even with less active features. In such settings, disconnection events can be avoided if deteriorating communication links are detected early so that prevention measures can be taken. In this paper we investigate how we can predict if a communication link will be terminated in the near future based on the recent trend of the signal. We propose a deep neural network framework which is executed onboard and we evaluate its performance based on simulation and real word data. The results show that we can predict the termination of a link up to 7 seconds into the future with 72.38% accuracy and 86.38% recall.