Leveraging AI for Dynamic Vehicle Control Systems

Abstract

Fueled by the rapid advancements in computational hardware, sensor technology, and machine learning, AI is increasingly adopted for a wide range of vehicle functionalities. Seemingly ever-increasing levels of autonomy, advanced driver assistance systems, and predictive maintenance are entirely built upon AI methods. The automotive sector has seen a considerable history of AI and machine learning applications in the last two decades. A notable point at which AI showed its potential as a transformative technology in automotive was in 2004 when a self-driving car demonstrated its ability to perceive its surroundings and make decisions in real time.

Objectives and challenges. Systems of an advanced AI vehicle must continuously and efficiently process spatiotemporal data to act in real time. In comparison to implementing a more conventional manually designed system, the system should autonomously process real-time data to act in real time. Thus, AI vehicle systems are faced with challenges across various fields, such as machine learning, optimization, and control, among others. In the expansive automotive sector, in its various design and operation aspects, numerous AI vehicles or vehicle systems have been developed. A remarkable example is the work on autonomous robotic systems and predictive maintenance. In addition to these examples, AI ship autopilot systems and highly automated AI systems for drones and other vehicles have additionally been proposed. Taken together, these examples highlight the automotive sector’s interest in self-optimizing AI systems. More recently, AI is increasingly used for adaptive vehicle control systems. For instance, AI has been shown to outperform rule-based predictive energy management systems, and drive cycles improved by AI-based gear shifting show increased accuracy.