Mirroring human behaviour and machine in autonomous cars

The thesis is a modest attempt to study the human-robot collaboration and how robots can learn from humans in the context of autonomous driving. The research was conducted in five distinct but interrelated phases. In the first part, we carry out a systematic review of driving simulators and cognitive architectures with an emphasis on ACT-R, discussing the role and impact of various cognitive abilities, decision-making processes, and driving behaviours on the overall driving task. This part serves as the groundwork for further research into autonomous vehicle systems. In the second part, we explore lateral control in autonomous vehicles using ACT-R cognitive architecture. Adopting a "look ahead" strategy, we focus on predictive steering for navigating diverse road patterns. Integrating ACT-R's cognitive processes with steering mechanisms, we develop a model that emulates advanced driving behaviours. This model uses preview points for steering adjustments, improving lane maintenance and maneuver execution. This blend of cognitive science with control technology marks a notable advancement in autonomous driving, enabling precise and adaptable navigation. The third part of this study revisits the driver gaze behaviour, incorporating real-time data collection in urban environments. We compare different gaze models, propose a new center point model, and analyze the drivers' gaze direction under varying driving scenarios. In the fourth part, we conduct an in-depth eye-tracking study comparing the driving performance of Spinal Cord Injury (SCI) individuals using hand controls with able-bodied individuals. The eye-tracking data provides vital insights into the visual attention and driving behaviour of these two groups. Lastly, we propose a Machine Intelligence Quotient (MIQ) framework for evaluating the intelligence of autonomous vehicles. The MIQ framework consists of various factors and sub-factors of intelligence, presenting an innovative and standardized way of quantifying the 'intelligence' of autonomous vehicles.