استقرار یک سیستم توزیع کامل برای بهبود جریان های ترافیک شهری : A تجزیه و تحلیل عملکرد مبتنی بر شبیه سازی

Distributed, cooperative systems dedicated to road traffic self-organization are very attractive, but present some drawbacks. In particular, their cooperative nature makes them fairly inefficient when working with a reduced number of partners. This situation typically corresponds to the deployment stage, during which only a few vehicles cooperate. This time period cannot be avoided when pushing a new system to the market. We are interested in two features that are important for this kind of system: traffic jam detection and traffic alert transmission. For the first feature, we present a theoretical model that anticipates the proportion of equipped vehicles that allows an acceptable level of traffic jam detection, and we validate this model by simulation. For the second feature, we examine two ways to improve the system behaviour when the proportion of equipped vehicles is very low; their efficiency is tested through simulation. This study is innovative because the simulation platform we developed can take into account the directional behaviour of wireless communications in urban context for a low computational cost.

The development of urban, suburban or inter-urban transportation generates many traffic jams arising from structural reasons or unexpected events. One possible means of combating these traffic jams is to use a Personal Navigation Assistant (PNA). To allow the users to reorganize their trips, these PNA must be able to perform a dynamic routing based on up-to-date data provided by a Traffic Information System (TIS). Such systems have already been proposed. Most of them are academic projects (e.g., Notice [1], SaveTime [2], Street Smart [3], TraffCon [4], SOTIS [5], Cartel [6]); a few others are commercial products (e.g., TMC [7], Dash Express [8] or TomTom HD Traffic [9]). Whatever the approach, up-to-date traffic data must be provided to vehicles. Thus, it is necessary to measure the current state of traffic (measurement), construct a synthetic picture of this state (aggregation), and send this information to the vehicles concerned by dynamic routing (diffusion). Each of these three functions can be either centralized (i.e., performed by a common equipment) or distributed (i.e., performed by the vehicles). In this article, we suppose that all the above functions are distributed, leading to the definition of Distributed Traffic Information Systems (DTISs). DTIS are inherently cooperative, which means that they cannot be effective when used by too few vehicles. This feature makes introducing these systems on the market difficult because their early users might gain no advantage. This situation typically arises during the deployment stage, which cannot be avoided. In addition, too little research has been done to address this major problem affecting DTIS. The aim of this article is to study the efficiency of DTIS when they are used by a low proportion of road users, and examine the ways to make them efficient enough in this situation. We focus here on the measurement and diffusion tasks because they provide key data for the DTIS and they are directly impacted by a low number of cooperative contributors.

Distributed Traffic Information Systems (DTISs) are not efficient during their deployment because they need a sufficient number of partners to work properly. To convince potential early users, DTIS must provide a minimum initial level of service. Two aspects are essential in DTIS: the distributed detection of traffic perturbations and the transmission of traffic alerts. The originality of this paper is to study them as a function of the Proportion of Equipped Vehicle (PEV). This study is based on a simple cooperative model, which has been proven effective when most of the vehicles are equipped. It is supported by an original simulator that takes into account both data and vehicle traffics. We proposed and validated a traffic jam detection model, which indicates the minimum proportion of equipped vehicles required to provide an acceptable level of traffic jam detection, for a given traffic condition. The traffic alert transmission protocol used is intentionally very simple, so the results presented can be viewed as the lower bound of the potential transmission range. We showed that the system cannot work properly by itself for very low PEV (e.g., about 2% in our experiments). For this reason, the very first deployment must be based on an auxiliary transmission network. A low-cost way to reach this goal is to provide a distributed infrastructure composed of roadside relay stations (RRSs), which help to diffuse data. Our simulations show that a preliminary analysis of standard traffic conditions makes it possible to optimize the RRS locations. In particular, positioning RRS at the crossroads provides a clear improvement of transmission ranges for a given number of RRS, without any additional cost.