Are you interested in the sophisticated application of computer vision, machine learning, and deep learning approaches to enhance cooperation between an intelligent vehicle and its driver? In this book, the authors describe methods and implementations for safely taking control of a vehicle when the automatic system requires it.
In 2017, the Society of Automotive Engineers (SAE) determined six levels ranging from purely manual driving to fully automated driving. The authors give solutions for the current level (3) of driving automation, which is conditional driving automation: “For certain driving scenarios, the vehicle could achieve the entire DDT and monitor the driving environment with the expectation that the driver executes the appropriate response for the requests to intervene triggered by system failure.”
The problem is that during a longer period of automatic operation the driver deals with non-driving-related activities (NDRAs) such as emailing, playing, or reading, and thus cannot immediately take control, either for physiological or psychological reasons. The authors explain: “monitoring the drivers’ state and activities that they are engaging in is crucial for the design of a smart human–machine interface [HMI] to improve their situation awareness before the takeover process.” “Safe, smooth, and swift control authority transitions between a human driver and the automated functionality of the vehicle” are crucial. To minimize the risks of a takeover, the authors give solutions for the system to recognize the state of the person.
The first part of the book describes ways for characterizing the status of the driver via computer vision tools: evaluation of eye gaze, hand gesture, head movement, the fusion of head and hand movements, and facial information via artificial intelligence (AI) algorithms.
Based on this, later chapters explore real-time driver behavior recognition, the implication of non-driving tasks on the takeover process, and driver workload estimation. Chapter 6 proposes a novel lightweight temporal attention-based convolutional neural network (LTA-CNN) “to achieve slow latency inference on the on-vehicle edge device in video-based NDRA recognition.” Chapter 7 discusses the implication of NDRAs on the takeover process. The next chapter focuses on driver workload estimation.
After detailing some specific aspects and their solutions, the last chapters give a wider perspective on the collaboration between human driver and vehicle automation: “Mutual understanding, as a critical aspect for multi-agent teaming and collaboration, enables the human driver and vehicle automation to collaborate efficiently by understanding the capability, intention, and attitudes of the teammate.” This reduces decision-making latency.
Chapter 9 deals with the characterization of neuromuscular dynamics for HMIs. The following chapter applies neuromuscular measuring dynamics for the introduction of a driver steering intention prediction system; “shared control-based methods were preferable to those with the immediate shutdown of automation.” The last chapter discusses an “intelligent haptic interface design for human–machine interaction in automated vehicles.”
The book and its themes are well arranged and detailed. Each chapter starts with an introduction/motivation part, references to related work, and a description of the model and method, and ends with results, analyses, conclusions, future work, and a rich list of references. The internal sections of the chapters are suited to their special theme. The book has a global index.
The authors are proficient specialists in their research areas, including computer vision, image processing, deep learning, machine learning, human behavior analysis, automated driving and human-machine systems, computing/simulation and modeling, industrial automation, instrumentation, sensors and measurement, systems engineering, through-life engineering services, and nondestructive testing and evaluation in digital manufacturing.
This book is for researchers working on developing safe and efficient human-machine collaboration, and for graduate students interested in these methods and sophisticated solutions.