PhD Oral Exam - Behrouz Samieiyan, Information and Systems Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

Autonomous vehicles (AVs) face a foundational trade-off between capability and energy. The sensing, computation, and communication capabilities that enable safe autonomous operation impose energy demands that onboard powertrains, particularly conventional internal combustion engine (ICE) and hybrid systems, struggle to sustain. The dissertation addresses this trade-off by treating traffic forecasting and energy management as coupled problems, structured around four research questions. First, a systematic literature review of energy management in autonomous vehicles (2018–2026) maps existing rule-based, heuristic, optimization-based, and AI-driven methods, identifying fragmentation between traffic-prediction and energy-management research as a central gap. Second, PhysGT, a physics-motivated graph-temporal network combining diffusion graph convolution with a novel FD-CrossAttention module, is proposed for joint traffic flow and speed prediction, achieving statistically stable improvements over graph-based baselines across three benchmark datasets. Third, three complementary energy management frameworks for a single conventional AV are developed and compared: ANFIS-TLBO-MPC, ANFIS-QIFA-MPC, and a fully online, learning-based CQ-RGNC policy, evaluated across four standardized highway driving cycles for fuel consumption and emissions. Fourth, a plug-and-play predictive cooperative framework for gasoline-powered AV platooning, combining ANFIS-TLBO powertrain modeling, Model Predictive Control, and QLSTM-based leader-state prediction, is shown to achieve the lowest total platoon fuel consumption across all tested cycles and platoon sizes without requiring modification to local vehicle controllers. Together, these contributions demonstrate that anticipatory, prediction-informed strategies consistently outperform reactive alternatives, and establish a conceptual and methodological foundation for integrating traffic-level forecasting with vehicle- and fleet-level energy management in autonomous mobility.