Название: From AI to Autonomous and Connected Vehicles: Advanced Driver-Assistance Systems (ADAS), Volume 2
Автор: Abdelaziz Bensrhair, Thierry Bapin
Издательство: Wiley-ISTE
Год: 2021
Страниц: 288
Язык: английский
Формат: pdf (true)
Размер: 36.67 MB

The main topic of this book is the recent development of on-board advanced driver-assistance systems (ADAS), which we can already tell will eventually contribute to the autonomous and connected vehicles of tomorrow.

With the development of automated mobility, it becomes necessary to design a series of modules which, from the data produced by on-board or remote information sources, will enable the construction of a completely automated driving system. These modules are perception, decision and action. State-of-the-art AI techniques and their potential applications in the field of autonomous vehicles are described.

Perception systems, focusing on visual sensors, the decision module and the prototyping, testing and evaluation of ADAS systems are all presented for effective implementation on autonomous and connected vehicles.

Communications, a requirement for developing far away perception and anticipating problems. These means of communication are called VANETs (Vehicular Ad-hoc NETwork) and follow the 802.11p standard. Different standards are available to manage these communication media dedicated to transport systems, for example, IEEE which proposes the WAVE architecture (Wireless Access in Vehicular Environments), ETSI which proposes ETSI TC-ITS architecture, and ISO offers the CALM architecture. In all cases, the goal is to provide media with a high Quality of Service (QoS). This QoS must guarantee the capacity of the communication medium to transmit data under the best possible conditions while respecting the criteria of availability, flow, transmission delays and minimum rate of packet loss (messages).

In the context of the development and deployment of automated vehicles, the use of communications is becoming a clearly essential and critical issue. Indeed, the information will be used to feed active applications (emergency braking, application of collision avoidance maneuver, vehicle platoon stability). In this context, the information transmitted by the means of communication will make it possible to update local dynamic perception maps, and also obtain extended dynamic perception maps (in range and in attributes). This extended perception makes it possible to feed the decision-making, path planning and action systems required for driving automation. The act of extending perception is also useful for anticipating and predicting future risky situations, as well as for generating optimal decisions (from a safety point of view) all the while controlling the maneuvers of automated vehicles.

Artificial Intelligence (AI) techniques may work very well without their output being easy to understand, even when performing an a posteriori audit. However, in certain areas involving discrimination between individuals (decision to award credits, for example), or other areas with strong safety stakes (driving an autonomous vehicle on an open road, for example), transparency requirements are demanding. In the first case, transparency or “explainability” makes it possible to justify a decision, whereas in the second case, it helps produce a formal security proof, for example by using the SIL (Security Integration Level) formalism. Those works which make the functioning of AI systems explainable are grouped under the term XAI (eXplainable Artificial Intelligence). One should note that – for this subject – knowledge-based systems have an advantage over purely numerical computing methods, such as Deep Learning, for instance. The dilemma when retrieving an XAI is that one generally must accommodate a decrease in performance in order to gain better explainability.

This book also addresses cooperative systems, such as pedestrian detection, as well as the legal issues in the use of autonomous vehicles in open environments.

Table of Contents:
Preface
1 Artificial Intelligence for Vehicles
1.1. What is AI?
1.2. The main methods of AI
1.3. Modern AI challenges for the industry
1.4. What is an “intelligent” vehicle?
1.5. References
2 Conventional Vision or Not: A Selection of Low-level Algorithms
2.1. Introduction
2.2. Vision sensors
2.3. Vision algorithms
2.4. Conclusion
2.5. References
3 Automated Driving, a Question of Trajectory Planning
4 From Virtual to Real, How to Prototype, Test, Evaluate and Validate ADAS for the Automated and Connected Vehicle?
5 Standards for Cooperative Intelligent Transport Systems (C-ITS)
6 The Integration of Pedestrian Orientation for the Benefit of ADAS: A Moroccan Case Study
7 Autonomous Vehicle: What Legal Issues?