Ifsttar PhD subject

French version
Detailed form :

Title : Cooperative control for road traffic management.

Main host Laboratory - Referent Advisor   -

Director of the main host Laboratory   -

PhD Speciality Mathématiques appliquées

Axis of the performance contract 1 - COP2017 - Efficient transport and safe travel

Main location Marne-la-Vallée

Doctoral affiliation UNIVERSITE DE MARNE-LA-VALLEE

PhD school MATHEMATIQUES ET SCIENCES ET TECHNOLOGIES DE L'INFORMATION ET DE LA COMMUNICATION (MSTIC)

Planned PhD supervisor IMINE Hocine  -  Université Gustave Eiffel  -  COSYS - PICS-L

Planned financing Contrat doctoral  - Ifsttar

Abstract

Context
The development of new intelligent transport systems continues to mobilize a large number of researchers and laboratories over the world. This is due to the desired and expected gains from these new systems, particularly in terms of the traffic management and of the reduction of pollutant emissions. Although several levels of decision-making and traffic management are needed, the local (decentralized) level becomes very interesting with the increasing development of the vehicle automation. Indeed, vehicles continue to be equipped with different types of sensors, computing capacities and various communication equipments. However, the information is often local and the computational capacities as well as the communication ranges are limited. The cooperative control approach of multi-agent systems consists in determining the optimal behavioral laws of the different agents in feedback on the state of the system at the local level; the optimality being defined on the overall state of the system. Applied to the road traffic management, this would result in maximizing cooperation between communicating agents at the local level (neighboring agents) to achieve an overall objective in the road network (time spent, emissions, etc.).

Traffic modeling and management in transportation networks are among the key themes of the Grettia laboratory. Grettia has been working for several years now, on both the development of traffic models including the cooperation aspect between agents (vehicles, infrastructures, etc.), and on the development of traffic regulation strategies benefiting from the contribution of these cooperative transportation systems [10,11]. Lepsis laboratory has been contributing for several years to the development of methods and tools for driver assistance systems, which has given rise to various publications in conferences and international journals. The involvement of Lepsis in national projects, such as VIF (Interactive Vehicle of the Future) [7] and European, such as HEAVYROUTE ([8], [9]), has made it possible to validate in an experimental manner the methods for vehicle control, especially those concerning the heavy vehicle, to avoid the risks of accident (rollover, jackknifing, ...) and for the planning of its trajectory.

This Grettia-Lepsis collaboration will allow Lepsis to extend its work to focus not only on an isolated vehicle but also on road traffic. We note that this collaboration Grettia-Lepsis on this topic is not new, since in the framework of the project IRS (Intelligent Routing System), submitted to the H2020 call, of which Lepsis was coordinator and Grettia was a partner, we had already proposed to work in the field of cooperative road traffic control. The thesis proposed here has also strong links with the Scoop@f project, to which both Grettia and Lepsis contribute. On the other hand, this thesis fits perfectly with the theme of the COP's deliverable entitled "Nouveaux modèles de trafic permettant la gestion de la mobilité coopérative" (New traffic models for the management of cooperative mobility) being prepared by Nadir Farhi at Grettia. We also emphasize that this proposal has a strong link with the "Mobilité connectée" (Connected mobility) axis of the “Projet Fédérateur” (Ifsttar unifying project) entitled "Mobilités et transitions numériques” (Mobilities and digital transitions).

Purpose
The purpose of this thesis is to investigate this cooperative control approach applied to the road traffic management, with the aim of proposing new models of cooperation including the modeling of traffic flows. We are interested here in the microscopic modeling of the traffic, but also and especially in progressing to the macroscopic modeling scale, which should make it possible to understand the effect of network-wide cooperation; and therefore, optimize the management of traffic on the network. The models should consider different penetration rates for the communication and different levels of vehicle automation.

Methodology
One of the most treated approaches of cooperative control is the development of the so-called “consensus algorithms”. When agents adhere to a certain value, they are said to have reached a consensus. Agents sharing information on the network have a consistent view of this information, which is critical to the coordination task. To reach consensus, agents share a variable of interest as well as negotiation algorithms to converge to consensus. As shared value, we can have the speed of vehicles, the inter-vehicular distance, a level of pollution, the shape of the platoon (multi-lane traffic), etc. Since vehicle interactions are assumed to be local, the shared algorithms are designed to be distributed. Agents update the value of the shared variable based on the traffic state at the local level. The objective here is then to define update rules such that the variable shared by all agents converge to the same value. The search for consensus is not obvious, especially because of the time-variant and dynamic characters of the communication topologies.

Scientific challenges
High efficiency and operational aptitude can be achieved by a group of agents operating in coordination. The main theoretical and practical challenges encountered in the development of cooperative control models [5] are 1) the development of a system of subsystems instead of a single system, 2) the communication bandwidth and connectivity are often limited, and the information exchanged between vehicles may be unreliable. It is also difficult to decide what, when and with whom to communicate. 3) the arbitration between individual and collective goals is negotiated. 4) the computing capacity of vehicles are often limited. The communication topology has an effect on the stability of the formation of groups of agents (vehicles), especially with limited v2v communication [1]. The change of the communication topology can also have an effect on group behaviors [2]. The innovative nature of this proposal compared to the state of the art is the scaling up of the cooperative traffic modeling and control to the macroscopic scale, while taking into account the various challenges mentioned above. Another scientific challenge for this thesis is the development of a cooperative control model allowing a good compromise between the modeling of road traffic dynamics, and the gains from the cooperation at the local level for an overall optimization on the road network.

Workplan
1. Literature review (from T0 to T0 + 6 months).
2. Model development for cooperative traffic management. (from T0 + 6 m. to T0 + 24 m.).
3. Numerical simulation (from T0 + 12 m. to T0 + 30 m.)
4. Writing of the thesis (from T0 + 24 m. to T0 + 36 m.)

Keywords : Cooperative control, Cooperative ITS, traffic modeling, traffic management

Main host Laboratory - Referent Advisor	-
Director of the main host Laboratory	-
PhD Speciality	Mathématiques appliquées
Axis of the performance contract	1 - COP2017 - Efficient transport and safe travel
Main location	Marne-la-Vallée
Doctoral affiliation	UNIVERSITE DE MARNE-LA-VALLEE
PhD school	MATHEMATIQUES ET SCIENCES ET TECHNOLOGIES DE L'INFORMATION ET DE LA COMMUNICATION (MSTIC)
Planned PhD supervisor	IMINE Hocine - Université Gustave Eiffel - COSYS - PICS-L
Planned financing	Contrat doctoral - Ifsttar
Abstract Context The development of new intelligent transport systems continues to mobilize a large number of researchers and laboratories over the world. This is due to the desired and expected gains from these new systems, particularly in terms of the traffic management and of the reduction of pollutant emissions. Although several levels of decision-making and traffic management are needed, the local (decentralized) level becomes very interesting with the increasing development of the vehicle automation. Indeed, vehicles continue to be equipped with different types of sensors, computing capacities and various communication equipments. However, the information is often local and the computational capacities as well as the communication ranges are limited. The cooperative control approach of multi-agent systems consists in determining the optimal behavioral laws of the different agents in feedback on the state of the system at the local level; the optimality being defined on the overall state of the system. Applied to the road traffic management, this would result in maximizing cooperation between communicating agents at the local level (neighboring agents) to achieve an overall objective in the road network (time spent, emissions, etc.). Traffic modeling and management in transportation networks are among the key themes of the Grettia laboratory. Grettia has been working for several years now, on both the development of traffic models including the cooperation aspect between agents (vehicles, infrastructures, etc.), and on the development of traffic regulation strategies benefiting from the contribution of these cooperative transportation systems [10,11]. Lepsis laboratory has been contributing for several years to the development of methods and tools for driver assistance systems, which has given rise to various publications in conferences and international journals. The involvement of Lepsis in national projects, such as VIF (Interactive Vehicle of the Future) [7] and European, such as HEAVYROUTE ([8], [9]), has made it possible to validate in an experimental manner the methods for vehicle control, especially those concerning the heavy vehicle, to avoid the risks of accident (rollover, jackknifing, ...) and for the planning of its trajectory. This Grettia-Lepsis collaboration will allow Lepsis to extend its work to focus not only on an isolated vehicle but also on road traffic. We note that this collaboration Grettia-Lepsis on this topic is not new, since in the framework of the project IRS (Intelligent Routing System), submitted to the H2020 call, of which Lepsis was coordinator and Grettia was a partner, we had already proposed to work in the field of cooperative road traffic control. The thesis proposed here has also strong links with the Scoop@f project, to which both Grettia and Lepsis contribute. On the other hand, this thesis fits perfectly with the theme of the COP's deliverable entitled "Nouveaux modèles de trafic permettant la gestion de la mobilité coopérative" (New traffic models for the management of cooperative mobility) being prepared by Nadir Farhi at Grettia. We also emphasize that this proposal has a strong link with the "Mobilité connectée" (Connected mobility) axis of the “Projet Fédérateur” (Ifsttar unifying project) entitled "Mobilités et transitions numériques” (Mobilities and digital transitions). Purpose The purpose of this thesis is to investigate this cooperative control approach applied to the road traffic management, with the aim of proposing new models of cooperation including the modeling of traffic flows. We are interested here in the microscopic modeling of the traffic, but also and especially in progressing to the macroscopic modeling scale, which should make it possible to understand the effect of network-wide cooperation; and therefore, optimize the management of traffic on the network. The models should consider different penetration rates for the communication and different levels of vehicle automation. Methodology One of the most treated approaches of cooperative control is the development of the so-called “consensus algorithms”. When agents adhere to a certain value, they are said to have reached a consensus. Agents sharing information on the network have a consistent view of this information, which is critical to the coordination task. To reach consensus, agents share a variable of interest as well as negotiation algorithms to converge to consensus. As shared value, we can have the speed of vehicles, the inter-vehicular distance, a level of pollution, the shape of the platoon (multi-lane traffic), etc. Since vehicle interactions are assumed to be local, the shared algorithms are designed to be distributed. Agents update the value of the shared variable based on the traffic state at the local level. The objective here is then to define update rules such that the variable shared by all agents converge to the same value. The search for consensus is not obvious, especially because of the time-variant and dynamic characters of the communication topologies. Scientific challenges High efficiency and operational aptitude can be achieved by a group of agents operating in coordination. The main theoretical and practical challenges encountered in the development of cooperative control models [5] are 1) the development of a system of subsystems instead of a single system, 2) the communication bandwidth and connectivity are often limited, and the information exchanged between vehicles may be unreliable. It is also difficult to decide what, when and with whom to communicate. 3) the arbitration between individual and collective goals is negotiated. 4) the computing capacity of vehicles are often limited. The communication topology has an effect on the stability of the formation of groups of agents (vehicles), especially with limited v2v communication [1]. The change of the communication topology can also have an effect on group behaviors [2]. The innovative nature of this proposal compared to the state of the art is the scaling up of the cooperative traffic modeling and control to the macroscopic scale, while taking into account the various challenges mentioned above. Another scientific challenge for this thesis is the development of a cooperative control model allowing a good compromise between the modeling of road traffic dynamics, and the gains from the cooperation at the local level for an overall optimization on the road network. Workplan 1. Literature review (from T0 to T0 + 6 months). 2. Model development for cooperative traffic management. (from T0 + 6 m. to T0 + 24 m.). 3. Numerical simulation (from T0 + 12 m. to T0 + 30 m.) 4. Writing of the thesis (from T0 + 24 m. to T0 + 36 m.)
Keywords :	Cooperative control, Cooperative ITS, traffic modeling, traffic management

List of topics

Applications closed