In this study, the problem of car-following stability is considered. This study presents a clustered car-following strategy based on inter-vehicle communications (IVCs), which contributes to derive a more accurate feedback control for improving the stability. In order to maintain car-following stability, two conditions are considered, which are based on the control theory and a modified optimal velocity car-following model. In simulations, the results demonstrate that the strategy is effective on the stability of car-following, and the effect becomes more obvious as the market penetration rates of IVC-equipped vehicles increase. Even in low market penetration rate (5%), the proposed strategy has significant influence on the stability and the recovering time.

Smartphone applications became very popular nowadays as they provide useful functionalities to our daily lives over ordinary voice services. They offer a small but powerful computing platform where intelligent algorithms can be coded into some live-saving products which profoundly impact our daily lifestyles. A mobile fatigue detection system which is proposed in this study is an important life-saving application running on smartphone. Mobile detection system faces technological challenges such as: (i) embracing relatively low-resolution images in image recognition; (ii) supporting fast response time given a low-power CPU in comparison to a desktop computer; and (iii) demanding for high prediction accuracy by light-weight machine learning algorithms, as the software programs embedded in smartphones is resource constrained. Solutions with respect to indoor facial profiling system which are mainly based on progressive locating method for eye detection are discussed in this study. Acceptable experimental results in terms of eye detection rate and driver fatigue detection in different situations are presented.

Computer vision methods for advanced driver assistance systems (ADAS) must be developed considering the strong requirements imposed by the industry, including real-time performance in low cost and low consumption hardware (HW), and rapid time to market. These two apparently contradictory requirements create the necessity of adopting careful development methodologies. In this study the authors review existing approaches and describe the methodology to optimise computer vision applications without incurring in costly code optimisation or migration into special HW. This approach is exemplified on the improvements achieved on the successive re-designs of vehicle detection algorithms for monocular systems. In the experiments the authors observed a ×15 speed up between the first and fourth prototypes, progressively optimised using the proposed methodology from the very first naive approach to a fine-tuned algorithm.

Vehicular control delay is important in evaluating the level of service and designing or adjusting the control schemes at signalised intersections. Although control delay may be obtained directly or indirectly nowadays for vehicles with global positioning system (GPS) or at locations with designated sensor types and configurations, such data are still not widely available. The analytical and empirical procedures for control delay estimation are still important, particularly in developing countries. However, very few studies explored the effectiveness of existing control delay estimation methods. This study intends to address this issue by comparing the delays estimated by four commonly adopted models using field data observed from three signalised intersections in Shanghai. Specifically, the study compares the delay estimates from deterministic queuing model, Webster model, highway capacity manual (HCM) 2000 model, and Shanghai adjusted model. The results indicate that HCM 2000 model and Shanghai adjusted model perform similarly and satisfactorily under various v/c ratios while deterministic queuing model has better performance when the v/c ratio is extremely high, but Webster model is mostly inadequate at coordinated signalised intersections. Meanwhile, signal progression, initial queue, proportion of vehicles arriving on green are found as main influential factors to the precision of models based on data analysis.

Traffic signal control at roundabouts is proved to be effective in addressing heavy traffic. However, the applicability of a signalised roundabout, especially compared with a signalised intersection, in terms of capacity has not been investigated. For signalised roundabouts, unsignalised roundabout capacity models including gap-acceptance theories and regression models are not applicable. This study first introduces a shockwave-based model to capture the impacts of signals at circulatory lanes on queuing and clearance time of left-turning vehicles. On this basis, a general signal timing model and a capacity model are developed. A sensitivity analysis is then elaborated to test how temporal and spatial parameters affect the capacity of a signalised roundabout. Capacity comparisons are made among signalised roundabouts, unsignalised roundabouts and signalised intersections. The results show that (i) under high traffic demand, a signalised roundabout outperforms an unsignalised roundabout; (ii) a signalised roundabout achieve higher capacity than a signalised intersection does under certain condition, (iii) but a signalised intersection outperforms a signalised roundabout when optimal cycle length is relatively short (65 seconds in this study), (iv) capacity of a signalised roundabout is very sensitive to the variation of signal parameters, hence care should be taken to design the parameters to ensure high capacity.

This study presents the adaptive traffic signal control algorithm in a distributed traffic network system. The proposed algorithm is based on a micro-simulation model and a reinforcement learning method, namely approximate dynamic programming (ADP). By considering traffic environment in discrete time, the microscopic traffic dynamic model is built. In particular, the authors explore a vehicle-following model using cellular automata theory. This vehicle-following model theoretically contributes to traffic network loading environment in an accessible way. To make the network coordinated, tunable state with weights of queue length and vehicles on lane is considered. The intersection can share information with each other in this state representation and make a joint action for intersection coordination. Moreover, the traffic signal control algorithm based on ADP method performs quite well in different performance measures witnessed by simulations. By comparing with other control methods, experimental results present that the proposed algorithm could be a potential candidate in an application of traffic network control system.

Existing in-vehicle sensing systems are concentrated on obstacle detection for pedestrian or vehicle. Limited work has been conducted on multi-class obstacle classification. This study addresses on this issue and aims to develop an approach for simultaneous detection and classification of multi-class obstacles. Stereovision is first used to segment obstacles from traffic background, then an improved active contour model is adopted to extract complete contour curve of the detected obstacles. Based on the contour extracted, geometrical features including aspect ratio, area ratio and height are integrated for classifying object types including vehicle, pedestrian and other obstacles. The approach was tested on substantial urban traffic images and the corresponding results prove the effectiveness of the proposed approach.

Overtaking manoeuvre is a key issue for two-lane rural roads. These roads should provide sufficient overtaking sight distance (OSD) at certain locations to allow faster vehicles to pass slower ones. However, overtaking requires occupying the opposing lane, which represents a serious safety concern. Severity of overtaking related crashes is very high, compared with other manoeuvres. The development of advanced driver assistance systems (ADAS) for overtaking is being a complex task. Only few systems have been developed, but are not still in use. This research incorporated accurate data of real manoeuvres to improve the knowledge of the phenomenon. The trajectory of the overtaking vehicles on the left lane was observed. An instrumented vehicle measured the overtaking time and distance, the abreast position, and the initial and final speed of 180 drivers that passed it during a field experiment. Six different kinematic models (such as uniform acceleration or linear variation of acceleration) were calibrated. Generally, drivers started to accelerate before changing to the opposing lane. These models may be applied to ADAS, to estimate OSD and to improve microsimulation models.