Search

As automated vehicles (AVs) become advanced, there is a growing concern over how AVs should interact with pedestrians. Increasing attention has, therefore, been drawn to pedestrian crossing behaviour research. Given the complexity of human behaviour and the traffic environment, existing studies have identified many influential factors related to pedestrian crossing behaviour. An important problem, however, is the need for more effort to uncover the human psychological mechanisms underpinning these observed behavioural patterns. Hence, the key aim of this project is: to narrow the gap between psychology and pedestrian crossing behaviour by bringing ideas from psychology into the analysis of pedestrian crossing behaviour and modelling this behaviour from a psychological perspective. This doctoral project conducted a range of research, including experimental study and empirical data analyses, to investigate pedestrian crossing behaviour in different traffic scenarios, i.e., uncontrolled intersections with a constant-speed vehicle, constant-speed continuous traffic flow, or a yielding vehicle. It was found that visual looming theta_dot (the rate of change of the optical size of the vehicle on the pedestrian's retina) is significantly negatively related to the percentage of crossing gap acceptance in constant-speed scenarios, supporting that looming may cause a sense of collision threat that affects pedestrian crossing decisions. In vehicle-yielding scenarios, the empirical data indicated that another looming-related visual cue tau_dot (the rate of change of tau, tau=theta/theta_dot) is a potential visual cue for detecting vehicle-yielding behaviour. A hybrid perception framework was then developed to account for pedestrian crossing behaviour by combining both theta_dot and tau_dot. In continuous constant-speed traffic flow scenarios, it was found that pedestrians might dynamically adjust their crossing decisions by comparing theta_dot of the previously rejected gap, the currently faced gap, and the following gap. Based on these findings, this project developed models to characterise both pedestrian crossing decision and its time-dynamic nature. Crucially, validations across different datasets demonstrated that these models reproduce pedestrian crossing decisions qualitatively and quantitatively. Predictions from these models highlight the notion that looming-related visual cues are directly available to the pedestrian visual system. Finally, in addition to these psychological mechanisms and models, this project also provided novel observations in pedestrian crossing behaviour. It suggested that the behaviour of pedestrians tending to accept smaller gaps at higher vehicle speed conditions might lead to potential safety issues for pedestrians. Distracted pedestrians might self-regulate their engagement between the crossing task and distraction based on the traffic situation in the continuous traffic flow, such as time gap size. Moreover, in a vehicle behaviour estimation study, it was found that in the early stage of road-crossing scenarios, pedestrians tended to interpret low driving speeds as a signal to give way, regardless of whether the vehicle was slowing down. Overall, understanding pedestrian road-crossing behaviour and its underlying mechanisms is a difficult challenge. Beyond purely experimental research and data analysis, this project demonstrates that applying theories and models developed in psychology will bring considerable benefits to pedestrian road-crossing behaviour research.

The development of extremely-constrained environments having sensitive nodes such as RFID tags and nano-sensors necessitates the use of lightweight block ciphers. Indeed, lightweight block ciphers are essential for providing low-cost confidentiality to such applications. Nevertheless, providing the required security properties does not guarantee their reliability and hardware assurance when the architectures are prone to natural and malicious faults. In this thesis, considering false-alarm resistivity, error detection schemes for the lightweight block ciphers are proposed with the case study of XTEA (eXtended TEA). We note that lightweight block ciphers might be better suited for low-resource environments compared to the Advanced Encryption Standard, providing low complexity and power consumption. To the best of the author's knowledge, there has been no error detection scheme presented in the literature for the XTEA to date. Three different error detection approaches are presented and according to our fault-injection simulations for benchmarking the effectiveness of the proposed schemes, high error coverage is derived. Finally, field-programmable gate array (FPGA) implementations of these proposed error detection structures are presented to assess their efficiency and overhead. The proposed error detection architectures are capable of increasing the reliability of the implementations of this lightweight block cipher. The schemes presented can also be applied to lightweight hash functions with similar structures, making the presented schemes suitable for providing reliability to their lightweight security-constrained hardware implementations.