Grant: Discovery Grants Program - Individual

The research on future intelligent transportation systems is monopolar either on developing CAVs for enhanced navigation, passenger comfort, and safety or realizing powertrain electrification to reduce fuel consumption and exhaust emissions. While ensuring quality HQP training with multidisciplinary research emphasis, the overarching objective of proposed research is to build knowledge on developing new methods and system designs to predict human driving style at various automation levels to help vehicles make human-like decisions during critical scenarios and improve the energy management of autonomous electric vehicles (AEVs). According to the taxonomy established by the Society of Automotive Engineers (SAE), human interventions are deemed essential in automated driving for vehicles at levels 3-4. Therefore, a correct cognition of drivers' behavioral and cognitive intelligence and the capacity to interpret anthropomorphic decision-making is essential in realizing autonomous driving and the coexistence of automated and conventional road vehicles. The mixed traffic situation is common in urban driving when several categories of vehicles share the exact width of the carriageway without any physical segregation between motorized and non-motorized vehicles and poor lane discipline. In such traffic scenarios, developing a reliable human recognition model with a human-center-aware decision-making unit becomes more critical. This research intends to build a new system for driverless vehicles in multi-level autonomy levels by combining data-driven intelligent deep learning CNN for feature extraction and a probabilistic model as the decision-making unit in a hierarchical architecture. For the highly complex mixed-traffic situation, a novel upper-level intelligent driver recognition model is proposed based on improved deep reinforcement learning (DRL) by combining reinforcement learning (RL) and deep learning (DL). The proposed frameworks are developed, verified, and tested based on data obtained from a driving simulator tailored for autonomous driving. According to the US Department of Energy, electric vehicles (EVs) are 60-73% efficient, depending upon their drive cycle. In the case of energy harvesting from vehicular components, energy savings can exceed 77% and no tailpipe pollutants. This study develops safe and optimal eco-driving strategies for autonomous EVs via advanced rule-based and learning-based control algorithms using updated traffic flow information. For the hybrid EVs, the proposed study incorporates developing a new energy management strategy through lower-level speed control and optimal upper-level switching between various power sources. This proposal's impact will be on developing new technologies for connected and autonomous electric vehicles (CAEVs). It will promote transportation efficiency and safety, reduce road casualties, increase user acceptance of CAEVs, and help Canada's accountability towards net-zero emissions by 2050.