Existing autonomous driving systems of intelligent vehicles such as advanced driver assistant systems (ADAS) assess and quantify the level of potential safety threats. However, they may not be able to plan the best response to unexpected dangerous situations and do not have the ability to cope with uncertainties since not all vehicles can always keep a safe gap from preceding vehicles and drive at a desired velocity. Previous research has not taken such uncertainties into account, it is, therefore, necessary to develop models which are not restricted by the predefined movement patterns of a vehicle. Existing systems are based on a model that estimates the threat level based only on one factor Time-To-Collision (TTC). This approach is limited since it cannot handle all scenarios and ignores all uncertainties. To overcome these limitations, this paper utilised deep learning to develop a range of models that rely on a group of factors to reliably estimate the threat level and predict conflicts under uncertainty using the concept of looming '. Comparative analyses were undertaken by incorporating new varying input factors to each model (e.g., surrogate safety measures, vehicle kinematics, macroscopic traffic data). Real-world experiments demonstrated that adding new factors increases the reliability and sensitivity of the models. Results also indicated that the models that consider looming provide low false alarm rate extending their applications for a wider spectrum of traffic scenarios. This is paramount for ADAS as uncertainties are inherent in the deployment of connected and autonomous vehicles in a mixed traffic stream.

With the ever-increasing advancements in the technology of driver assistant systems, there is a need for a comprehensive way to identify traffic conflicts to avoid collisions. Although significant research efforts have been devoted to traffic conflict techniques applied for junctions, there is dearth of research on these methods for motorways. This paper presents the validation of a traffic conflict prediction algorithm applied to a motorway scenario in a simulated environment. An automatic video analysis system was developed to identify lane change and rear-end conflicts as ground truth. Using these conflicts, the prediction ability of the traffic conflict technique was validated in an integrated simulation framework. This framework consisted of a sub-microscopic simulator, which provided an appropriate testbed to accurately simulate the components of an intelligent vehicle, and a microscopic traffic simulator able to generate the surrounding traffic. Results from this framework show that for a 10% false alarm rate, approximately 80% and 73% of rear-end and lane change conflicts were accurately predicted, respectively. Despite the fact that the algorithm was not trained using the virtual data, the sensitivity was high. This highlights the transferability of the algorithm to similar road networks, providing a benchmark for the identification of traffic conflict and a relevant step for developing safety management strategies for autonomous vehicles.

As the automotive industry moves towards Autonomous Vehicles (AVs), developing reliable sensing systems such as lane marking detection, is crucial. Lane markings offer essential spatial and navigational cues for AVs' safety and efficiency. Therefore, it is vital to thoroughly understand the evolution and effectiveness of different lane marking detection technologies, particularly in identifying the challenges and influencing factors for their successful implementation. To facilitate an objective comparison, a comprehensive dataset was collected from a motorway in the UK employing an instrumented vehicle. This dataset contains representative scenarios including optimal conditions, faded lane markings, adverse weather conditions, nighttime and traffic congestion. Using this dataset, a comparative analysis of four prominent lane marking detection methods such as (i) Spatial Convolutional Neural Network, (ii) Lane detection model with attention mechanism, (iii) Inverse Saliency Pyramid Reconstruction Network (InSPyReNet) and (iv) REcurrent Feature-Shift Aggregator was conducted. InSPyReNet technique emerged superior, demonstrating outstanding precision and sensitivity in lane detection. The consistent evaluation approach in this research contributes to identifying the most suitable technique for robust lane marking detection under various environmental conditions. The study also navigates future research paths, emphasising model generalisation, and the importance of robustness in challenging conditions. The uniqueness of this research lies in its inclusive comparison of lane detection methods under varying conditions, evaluated on a single dataset. This not only serves as a valuable reference, but also opens new possibilities and avenues for future advancements in the field.