Axel Wegener

VANET Simulation Environment with Feedback Loop and its Application to Traffic Light Assistance

Traffic applications, in which vehicles are equipped with a radio interface and communicate directly with each other and the road traffic infrastructure are a promising field for ad-hoc network technology. Vehicular applications reach from entertainment to traffic information systems, including safety aspects where warning messages can inform drivers about dangerous situations in advance. As performance tests of the real system are very expensive and not comprehensive, todays evaluations are based on analysis and simulation via traffic simulators. In order to investigate the impact of traffic information systems there are two options: First, traffic simulators can be extended by application code and a simplified model for wireless communication. Second, existing network simulators can be coupled with existing traffic simulators. We favor the coupling of existing and well known simulators as we believe that the wireless communication characteristics influence the data transfer significantly and an oversimplified transmission model can lead to flawed results. In this paper we describe the feedback loop between traffic and network simulators named traffic control interface (TraCI) and outline its versatility. We explain its use to determine possible energy consumption reduction when traffic lights send their phase schedules to vehicles.

Hovering data clouds: a decentralized and self-organizing information system

With ever-increasing numbers of cars, traffic congestion on the roads is a very serious economic and environmental problem for our modern society. Existing technologies for traffic monitoring and management require stationary infrastructure. These approaches lack flexibility with respect to system deployment and unpredictable events (e.g., accidents). Moreover, the delivery of traffic reports from radio stations is imprecise and often outdated. In the project AutoNomos we aim at developing a decentralized system for traffic monitoring and managing, based on vehicular ad-hoc networks (VANETs). Our objective is to design a system for traffic forecasting that can deliver faster and more appropriate reactions to unpredictable events. In our design, cars collect traffic information, extract the relevant data, and generate traffic reports. A key concept are so-called Hovering Data Clouds (HDCs), which are based on the insight that many crucial structures in traffic (e.g., traffic jams) lead an existence that is independent of the individual cars they are composed of. The result is an elegant, robust and self-organizing distributed information system. In this paper we demonstrate first experimental results.

Empowered by wireless communication: Distributed methods for self-organizing traffic collectives

In recent years, tremendous progress has been made in understanding the dynamics of vehicle traffic flow and traffic congestion by interpreting traffic as a multiparticle system. This helps to explain the onset and persistence of many undesired phenomena, for example, traffic jams. It also reflects the apparent helplessness of drivers in traffic, who feel like passive particles that are pushed around by exterior forces; one of the crucial aspects is the inability to communicate and coordinate with other traffic participants. We present distributed methods for solving these fundamental problems, employing modern wireless, ad-hoc, multi-hop networks. The underlying idea is to use these capabilities as the basis for self-organizing methods for coordinating data collection and processing, recognizing traffic phenomena, and changing their structure by coordinated behavior. The overall objective is a multi-level approach that reaches from protocols for local wireless communication, data dissemination, pattern recognition, over hierarchical structuring and coordinated behavior, all the way to large-scale traffic regulation. In this article, we describe three types of results: (i) self-organizing and distributed methods for maintaining and collecting data (using our concept of Hovering Data Clouds); (ii) adaptive data dissemination for traffic information systems; (iii) methods for self-recognition of traffic jams. We conclude by describing higher-level aspects of our work.

Designing a Decentralized Traffic Information System — AutoNomos

We propose a decentralized traffic information system—Auto-Nomos—that is based on a thorough investigation of the properties of traffic and recommends a hierarchical data aggregation and forwarding for providing individualized information and support to road users. Our approach differs from work in the field by consequently applying local rules and local decentralized data processing, which turns out to be a key property of robust and scalable computing systems. We present a flexible VANET middleware that assists the application development by providing generic functionality for traffic applications. We discuss the architectural design of the overall system and provide solutions of important design concepts demonstrating the innovation of the approach.

Methods for Improving the Flow of Traffic

We describe a distributed and self-regulated approach for the self-organisation of a large system of many self-driven, mobile objects, i.e., cars in traffic. Based on methods for mobile ad-hoc networks using short-distance communication between vehicles, and ideas from distributed algorithms, we consider reactions to specific traffic structures (e.g., traffic jams.) Building on current models from traffic physics, we are able to develop strategies that significantly improve the flow of congested traffic. Results include fuel savings up to 40% for cars in stop-and-go traffic; we present a number of simulation results illustrating the underlying mechanisms. In addition, we discuss the organic structure of urban traffic, and hint at how self-healing methods can lead to improvements in rush-hour traffic.

GRAPE - Gradient based Routing for All PurposE

Routing within wireless mobile multi-hop networks is a challenging task. Therefore, in the past several routing protocols have been developed with various extensions and adoptions to certain specific scenarios. The consensus of the discussion about routing is that depending on all aspects of their application reaching from number of nodes, mobility, to the data traffic profile of the application developers can choose an optimal routing protocol. In the meantime some routing protocols have evolved as standard choice like AODV, DSR, OLSR, TBRPF. With the advent of sensor networks, requirements like scalability and resource consumption are added to the research domain. This resulted in additional extremely specialized routing algorithms or special extensions to existing protocols. So is there really a need and room for another routing protocol? We are convinced that a really simple, efficient, scalable protocol that allows for mobility and is suited for various performance classes of devices does not exist. We propose GRAPE as a completely new approach for routing in wireless multi-hop networks and present performance results based on simulations and measurements that demonstrate the potential of the routing approach.

Hovering Data Clouds for Organic Computing

As part of our project AutoNomos, we have investigated traffic information and management systems that motivate the usage of new methods and tools inspired by Organic Computing paradigms. Current traffic monitoring and management approaches with stationary infrastructure lack flexibility with respect to system deployment and have difficulties with detecting unpredictable events (e.g., accidents). One goal of AutoNomos is the development of a distributed and self-organising traffic information and management system without a centralised infrastructure. Our system relies on a GPS-based navigation system and a wireless radio interface; vehicles can gather information about the current position on the road network and form a vehicular ad-hoc network (VANET) to share information about traffic phenomena. In this article, we introduce Hovering Data Clouds (HDCs) as a tool to collect, aggregate and disseminate application-specific data. HDCs evolve in a self-organising manner at locations of relevant data in the system. Although their data is hosted on the nodes, HDCs exist independent of the individual carriers. While HDCs float between physical carriers, their corresponding HDC messages are disseminated in the network by a new effective transport protocol named AutoCast, designed according to Organic Computing paradigms. Finally, we demonstrate that HDCs detect traffic phenomena reliably and propagate them robustly within the network.