| Popis: |
There is a broad range of technologies available for wireless communications for moving vehicles, such as Worldwide Interoperability for Microwave Access (WiMax), 3G, Dedicated Short Range Communication (DSRC)/ Wireless Access for Vehicular Environment (WAVE) and Mobile Broadband Wireless Access (MBWA). These technologies are needed to support delay-sensitive safety related applications such as collision avoidance and emergency breaking. Among them, the IEEE802.11p standard (aka DSRC/WAVE), a Wi-Fi based medium RF range technology, is considered to be one of the best suited draft architectures for time-sensitive safety applications. In addition to safety applications, however, services of non-safety nature like electronic toll tax collection, infotainment and traffic control are also becoming important these days. To support delay-insensitive infotainment applications, the DSRC protocol suite also provides facilities to use Internet Protocols. The DSRC architecture actually consists of WAVE Short Messaging Protocol (WSMP) specifically formulated for realtime safety applications as well as the conventional transport layer protocols TCP/UDP for non-safety purposes. But the layer four protocol TCP was originally designed for reliable data delivery only over wired networks, and so the performance quality was not guaranteed for the wireless medium, especially in the highly unstable network topology engendered by fast moving vehicles. The vehicular wireless medium is inherently unreliable because of intermittent disconnections caused by moving vehicles, and in addition, it suffers from multi-path and fading phenomena (and a host of others) that greatly degrade the network performance. One of the TCP problems in the context of vehicular wireless network is that it interprets transmission errors as symptomatic of an incipient congestion situation and as a result, reduces the throughput deliberately by frequently invoking slow-start congestion control algorithms. Despite the availability of many congestion control mechanisms to address this problem, the conventional TCP continues to suffer from poor performance when deployed in the Vehicular Ad-hoc Network (VANET) environment. Moreover, the way non-safety applications, when pressed into service, will treat the existing delay-sensitive safety messaging applications and the way these two types of applications interact between them are not (well) understood, and therefore, in order for them to coexist, the implication and repercussion need to be examined closely. This is especially important as IEEE 802.11p standards are not designed keeping in view the issues TCP raises in relation to safety messages. This dissertation addresses the issues arising out of this situation and in particular confronts the congestion challenges thrown up in the context of heterogenous communication in VANET environment by proposing an innovative solution with two optimized congestion control algorithms. Extensive simulation studies conducted by the author shows that both these algorithms have improved TCP performance in terms of metrics like Packet Delivery Fraction (PDF), Packet Loss and End-to-End Delay (E2ED), and at the same time they encourage the non-safety TCP application to behave unobtrusively and cooperatively to a large extent with DSRC’s safety applications. The first algorithm, called vScalable-TCP – a modification of the existing TCPScalable variant – introduces a reliable transport protocol suitable for DSRC. In the proposed approach, whenever packets are discarded excessively due to congestion, the slow-start mechanism is purposely suppressed temporarily to avoid further congestion and packet loss. The crucial idea here is how to adjust and regulate the behaviour of vScalable-TCP in a way that the existing safety message flows are least disturbed. The simulation results confirm that the new vScalable-TCP provides better performance for real-time safety applications than TCP-Reno and other TCP variants considered in this thesis in terms of standard performance metrics. The second algorithm, named vLP-TCP – a modification of the existing TCP-LP variant – is designed to test and demonstrate that the strategy developed for vScalable-TCP is also compatible with another congestion control mechanism and achieves the same purpose. This expectation is borne out well by the simulation results. The same slow-start congestion management strategy has been employed but with only a few amendments. This modified algorithm also improves substantially the performance of basic safety management applications. The present work thus clearly confirms that both vScalable-TCP and vLP-TCP algorithms – the prefix ‘v’ to the names standing for ‘vehicular’ – outperform the existing unadorned TCP-Scalable and TCP-LP algorithms, in terms of standard performance metrics, while at the same time behaving in a friendly manner, by way of sharing bandwidth non-intrusively with DSRC safety applications. This paves the way for the smooth and harmonious coexistence of these two broad, clearly incompatible or complementary categories of applications – viz. time-sensitive safety applications and delay-tolerant infotainment applications – by narrowing down their apparent impedance or behavioural mismatch, when they are coerced to go hand in hand in a DSRC environment. |
