Deep Learning for Unmanned Autonomous Vehicles: A Comprehensive Review

Abstract

In recent years, deep learning as a subfield of machine learning has gained increasing attention due to its potential advantages in empowering autonomous systems with the ability to automatically learn underlying features in data at different levels of abstractions, to build complex concepts out of simpler ones and to get better with experience without being explicitly programmed. This book chapter provides a comprehensive review on the applications of deep learning in unmanned autonomous vehicles. We focus on particular research efforts that employ deep learning techniques to endow autonomous vehicles with different cognitive functionality, following the cognitive cycle of autonomous vehicles. This cognitive cycle of Sense-Aware-Decide-Act-Adapt-Learn extends the deliberative cycle of Sense-Decide-Act by adding situation awareness, adaptation and learning capabilities to autonomous vehicles. Potential applications of deep learning and major challenges are highlighted in this chapter.