Keywords

Adaptive Cruise Control, Optimization, Model Predictive Control, Reinforcement Learning, Electric Vehicle, vehicle-to-vehicle communication

Subject Categories

Automotive Engineering | Electrical and Electronics

Abstract

This dissertation develops information-driven methods to reduce traction energy in battery electric vehicles during adaptive and cooperative cruise control. Physics-grounded energetics are embedded in a predictive controller that accounts for intermittent V2V preview, sensing noise, packet loss, and powertrain limits. To ensure deployability, the nonconvex traction–power map is replaced by locally convex surrogates so each step solves a small, strictly convex QP in real time (average ≈ 70 ms/step on a desktop CPU: 8 cores/16 threads, 4.2–5.0 GHz), leaving margin at typical sampling rates (Ts =0.05–0.10 s; N=15–25).

Across standardized drive cycles from NREL DriveCAT—including FTP–75 (light duty), NREL Class 3 Electric, and Metro Highway (CA)—the energy-aware ACC (Strategy A) reduces battery energy relative to a conventional ACC baseline by an average of 2.30% over ten cycles (up to 12.61% on an arterial profle). With uncertainty-handling (Strategy B: dropout compensation, causal preview smoothing, and tightened headway constraints) and PER sweeps {0.1, 0.3, 0.5, 0.9}, savings remain robust and can increase on stop–go cycles (e.g., FTP–75 up to 15.8% with a median flter, W=15, constant-speed preview; NREL Class 3 up to 10.9%) while maintaining spacing safety and comfort bounds. The convex-surrogate MPC (Strategy C) closely tracks the nonlinear energy model and preserves these gains with predictable solve times. As a forward-looking extension, the dissertation also reports a completed deep reinforcement learning study for energy-aware ACC under packet loss, where SAC, TD3, and DDPG are compared and SAC shows the strongest energy–safety trade-off on urban, mixed, and highway drive cycles. All experiments were implemented in MATLAB/Python using cycle-driven simulations with headway h=0.8–1.2 s and standard acceleration/jerk limits. Overall, the work offers a reproducible, real-time pathway to energy-effcient longitudinal automation that degrades gracefully under packet loss while honoring powertrain envelopes.

Completion Date

2026

Semester

Summer

Committee Chair

Pourmohammadi Fallah, Yaser

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Electrical and Computer Engineering

Format

PDF

Document Type

Dissertation

Language

English

Download

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.