Machine Learning for Predictive Analytics in Autonomous Vehicle Fleet Management

Abstract

The increased interest and pilot deployment of autonomous vehicle technologies, in addition to the industrial and research investments, show the importance and impact of employing asset sharing and high automation technologies in logistics and last-mile operations. The breakthrough in machine learning and power-efficient computation techniques enables the use of sensors and systems that allow the deployment of self-driving cars. These vehicles can leverage multiple sensors to build digital maps, localize themselves, and estimate the traffic elements in their surroundings without missing or falling into autonomous position mismatching problems. Additionally, localization is complemented by sensors with wireless communication capabilities for stationary DGPS or RTK stations, in addition to CAN bus relative positioning techniques. Lastly, terrain and environmental adaptive security access measures can be instantiated through cooperative spectral algorithms. Although the exact mix of technologies is evolving, the use of predictive analytics and machine learning techniques in logistics and transport is presented in this paper.