Traffic congestion has become a major challenge for transportation professionals and roadway users across the world. As more of the world becomes more mobile, congestion during peak hours results in wasted time for billions of people around the globe. The effects of congestion delays on the individual are mostly negative: a reduction of air quality due to vehicle idling, drivers’ quality of life affected by having large amounts of non-productive time, which results in reduced time with family and friends, as well as economic losses due to non-productivity. Congestion also has a negative impact on safety, as it causes drivers to make increased decisions during stop and go traffic. Financial, environmental, and land use considerations provide an increasingly difficult environment to significantly increase the capacity of roadways by adding additional lanes. Fortunately, congestion can be alleviated by replacing human-operated vehicles with automated vehicles, which free the driver from the mental workload of a large number of tasks, some of which have to be carried out in parallel. The promise of reduced non-recurring congestion, due to reduction in vehicle crashes (approximately 25% of all congestion in the U.S.), provides great opportunities for the supplement of automated vehicles into the fleet. In addition, computer-operated vehicles have shorter reaction times, which allow the vehicles to be closer to one another, thus increasing traffic flow. Of all basic vehicular manoeuvres, lane changing is arguably one of the most difficult ones. There were approximately 539,000 two-vehicle lane change crashes United States alone in 1999. Analysis of the German In-Depth Accident Study from 1985 to 1999 shows that, on average, more than 5% of accidents occurred while changing lanes. In 2008, 1.7% of the registered highway accidents in the Netherlands were caused by inadequate lane changing. While it has been shown by Tsao et al. that the exit success percentage, which is the number of automated vehicles that successfully exit the system divided by the number of vehicles that need to exit, is well below 100% due to the lack of gaps sufficiently large enough for safe lane changes We believe that it is crucial to provide a mechanism that best utilizes available gaps. To achieve the promise of high throughput and increased safety, a technique that minimizes the disruption of traffic flow by automated vehicles during lane changes must be implemented. In this article, we are interested in designing such a technique with the objective of maximizing the safe number of possible lane changes. Although there exists a large number of automated lane change assistant systems, as shown in Section II, to the best of our knowledge, there has been no work that attempts to minimize the disruption of traffic flow by maximizing the number of lane changes for live traffic on a stretch of a highway with an arbitrary number of lanes, without any assumptions on vehicles’ dynamic attributes such as speeds. Our main contributions are as follows. * Given an arbitrary number of automated vehicles, we design an algorithm to maximize the number of possible lane changes on an arbitrary segment of a highway at any given time. The proposed algorithm uses information such as vehicles’ positions, speeds, and time slacks (to be defined later) to make judicious lane change decisions without requiring prior knowledge on traffic patterns or unnecessary braking. To reduce runtime overhead, we propose a distributed approach to allow for local lane changing decisions to be made during run time. * We present a lane change simulation platform that enables the implementation and comparison of different lane change algorithms. A large number of simulations can be run efficiently and various simulation parameters such as the number of vehicles wanting to change lanes can be specified. The remainder of the thesis is outlined as follows. We review existing literature regarding lane changes in Section 2. Section 2 also provides the system model and state the assumptions made in the thesis. The minimum time slack calculations, which is used to determine if a vehicle can change lanes without a collision, is presented in Section 3. Our distributed approach is discussed in Section 4 and the details of our online algorithm in Section 5. Section 5 also discusses the practical factors involved in implementing our approach in real operating scenarios. Simulation results presented in Section 6 and Section 7 conclude the thesis. (Author/publisher)
Abstract