The advent of new technologies in transportation, known as Intelligent Transportation Systems (ITS), and their introduction into daily city and intercity traffic has become a major preoccupation for car manufacturers, road authorities, traffic operators, and researchers. High technology is soughtafter in the current climate agitated by growing challenges related to emerging technologies. The exponential growth rate of smartphone use, the rising challenges of cloud computing and big data call for new solutions and at the same time bring new means to move towards those new solutions [1]. This technological context paves the way for the concept of connected roads, enabling infrastructure and vehicles to cooperate and exchange their data via various communication means thereby facilitating traffic monitoring in order to increase safety and efficiency, and decrease environmental impact. The diversity of data sources, including fixed cameras, infrared beacons, loop detectors, bluetooth detectors, GPS devices, drivers’ smartphones, vehicle radars and on board devices, provides many data sets and enables the establishment of sensor networks with unprecedented scalable monitoring capabilities.

The microscopic scale has always been an important area of interest and research as it directly affects the way people drive. Stakeholders and researchers have been insistently working for a future in which the human driving task would be considerably reduced or even removed, in order to minimize human errors. When Ralph Teetor thought of the first cruise control system which was later launched on a few 1958 Chrysler models, while traffic theory was still in its very early stages, the first step towards automated (driverless) and autonomous (driver present) vehicles was made. Autonoumous vehicles refer to the vehicles in which the driver is partially or fully assisted in his driving task by an electronic system, without any communication with other vehicles or infrastructure. At the same time, invehicle sensors, communication devices, and infrastructure units enable the frequent exchanges of information between vehicles and infrastructure and contribute to develop the connected vehicles, which designate vehicles with endowed communication capabilities. Cooperative vehicles are connected vehicles that share some of the functionalities of autonomous vehicles [2].

The constant progress in the implementation of Advanced Driver Assistance Systems (ADAS) has well expressed this paradigm shift towards mobility and safety. Now a multitude of ADAS systems [3] include infotainment, adaptive cruise control, lane changing assistant, night vision, automatic parking, among others, that aim at providing safer and more efficient traffic by supporting the driving task. A complete shift towards fully assisted vehicles was introduced in the last two decades, with the California Partners for Advanded Transportation TecHnology (PATH) pioneer work on platoons of automated vehicles, and very recently the autonomous Google driverless car. The industry’s move towards automation is mainly motivated by the perspective of a safer traffic where automated electronic systems prevent human mistakes and inefficiencies, as well as by the potential to reduce environmental impact. Prospective standardization efforts on ITS technologies has been made at an international level resulting in numerous funded Research and Development (R&D) projects, which give particular attention to the potential of ITS in decreasing hazardous situations and reducing greenhouse gas emissions. These emerging technologies also stimulate investment because they have the potential to make vehicles more marketable: where the technology is seen as fashionable it may attract customers and boost a company’s image.

In this context, a driver would legitimately claim to benefit from these cutting edge technologies. While the liability question in the case of collisions related to systems failures gives way to legal issues, especially in the case of fully automated vehicles, the safe behavior of ADAS and any integrated automated system must be a guarantee. Some open debates may then be raised with a system that potentially informs the driver, warns the driver or even takes over the driving task during identified critical situations:

• Does a partial support system offered to the driver lead to a less careful and less reactive driver?

• What is an appropriate design for the in-vehicle Human Machine Interface (HMI), as too much information may cause a lack of attention to the driving task?

• Are drivers ready to let an automated system drive for them, especially in critical situations?

• What happens in the case of a crash involving a normal vehicle and a cooperative vehicle?

• Is it possible to mix an equipped fleet of cooperative vehicles with normal vehicles without deteriorating the traffic flow characteristics?

Facing these concerns relating to the reality of a mixed traffic made of normal and cooperative vehicles, prospective simulation studies are much needed to counterbalance the lack of cooperative vehicles data and to figure out to which extent cooperative vehicles can help reduce traffic flow heterogeneities and collision-prone situations.

It is known that congestion wastes a massive amount of time, fuel and money. However congestion does not systematically come from a capacity default as about half of it is due to non-recurrent events [4], which are typically due to crashes, disabled vehicles, convoys of trucks or adverse weather conditions. Non-recurrent events follow irregular patterns in time, and their consequences on traffic flow efficiency and safety could therefore be dramatic [5]. A legitimate idea would consist in using the potentialities of cooperative vehicles to reduce the occurences of such non-recurrent events. The presented research takes on the problem of a mixed traffic on highways, in which cooperative vehicles have to interact with normal drivers, with an emphasis on the question of congestion appearance and spontaneous traffic jams, when shock waves appear “out of nowhere”. More specific challenges may then emerge regarding the use of cooperative vehicles for traffic flow management:

• What physics can explain the formation of non-recurrent traffic jams?
• What type of information, and how much information, can be utilized to control the dynamics of cooperative vehicles, and to reduce undesirable traffic situations?
• What control algorithms to integrate such data exchanges, and to prevent cooperative vehicles from adopting abnormal driving behavior?
• What traffic safety and efficiency indicators can well reflect traffic conditions?
• What is the proportion of cooperative vehicles needed to get positive effects?
• Can negative effects result from vehicle cooperation?
• How can the different sources of noise (drivers behavior, communication latency time, sensors faults) be integrated?

This dissertation aims at considering and addressing these questions, while the main intention is to operate cooperative vehicles to homogenize highway traffic flow thereby reducing the formation of spontaneous shock waves. First, an analytical framework was needed to understand the physics of traffic and the potential of cooperation, therefore, stability analyses of defined cooperative car-following models were conducted. Real data based calibration of car-following models was performed using a robust methodology in order to reproduce drivers behavior variability in simulation. Finally, a multi-agent modelling framework was introduced to consider all sources of perturbations, with the perspective of achieving self-organization in highway traffic.