Literature search on machine learning applied to autonomous vehicular driving: A review
Keywords:
Machine learning, Reinforcement learning, Deep learning, Artificial intelligence, Autonomous vehicles, Intelligent vehicles.Abstract
The aim of this paper is to explore the various industrial and research initiatives on the application of Artificial Intelligence (AI) techniques to autonomous vehicle driving. Particularly noteworthy in the study of autonomous vehicles (AV) is that the transition to an era of abundant data demands a paradigm shift from physics-based models to AI-driven methods capable of predicting future traffic dynamics and assisting in the formulation of optimized traffic policies, the potential of which lies in factors such as reducing human error and responding quickly to accidents in real time, factors that justify the study presented. Autonomous driving transcends traditional traffic patterns by performing tasks such as proactively recognizing critical events, planning next movements, making decisions and carrying out control tasks to ensure passenger safety and comfort in dynamic traffic environments. The levels of vehicle automation are presented and the focus is on AI-driven methods focused on End-to-End structures rather than pipeline structures, exploring details about MLP (Multi-Layer Perceptron) and KAN (Kolmogorov-Arnold Networks) Neural Network architectures, the main concepts and strategies that guide these techniques, as well as future challenges related to VAs. It can therefore be concluded that technologies such as machine learning, deep learning and reinforcement learning, as well as their combined use, are essential for the implementation of AV control systems that promote the evolution of the transportation system.
