Sensorless control for electric power steering in automotive applications: a systems engineering perspective

Abstract

With the rapid expansion of vehicle electrification and automated driving, Electric Power Steering (EPS) has become a foundational safety-critical subsystem in modern vehicles. Unlike hydraulic systems, EPS relies on electronic control of a Permanent Magnet Synchronous Motor (PMSM) to generate assist torque, enabling improved energy efficiency, packaging flexibility, and software-defined steering characteristics. As EPS systems increasingly operate under ISO 26262 (functional safety) and ISO/SAE 21434 (cybersecurity) constraints—particularly for ASIL-D applications—the architectural robustness of motor control becomes a system-level concern rather than a purely algorithmic one. Sensorless motor control has emerged as a promising approach to reduce dependency on hardware position sensors such as resolvers or Hall devices, thereby lowering cost, complexity, and failure exposure. This thesis investigates the structural feasibility of a hybrid sensorless control architecture for PMSM-based EPS, combining Sliding Mode Observer (SMO)–based back-EMF estimation in the medium-to-high speed region with High-Frequency Injection (HFI)–based saliency tracking for low-speed observability. Quantitative validation is performed in a MATLAB/Simulink environment using an IPMSM model with saliency. The results demonstrate stable and robust SMO-based estimation in the medium-to-high speed region, with consistent speed tracking and torque ripple characteristics under load variation. While full closed-loop low-speed hybrid validation remains future work, the HFI signal-processing path and architectural integration are implemented and structurally verified. Beyond algorithmic comparison, this study conducts an explicit EPS-oriented cross-evaluation of four major sensorless control strategies (EKF, SMO, MRAS, PLL) and establishes that the SMO–HFI hybrid offers the most favorable robustness-to-complexity tradeoff for sedan-class production EPS systems. The work adopts a systems engineering perspective, cascading vehicle-level safety, hardware, software, and cost constraints down to observer design. This traceability from system objectives to signal-level control structure constitutes the primary contribution of the thesis and defines a realistic pathway toward future steer-by-wire and higher-automation steering architectures. Most conventional studies on sensorless control focus narrowly on specific algorithmic techniques, such as SMO-based rotor position estimation or high-frequency signal injection methods, without explicitly addressing their integration into full vehicle steering architectures. In contrast, the present work adopts a systems engineering perspective grounded not only in academic research but also in practical development experience. Rather than optimizing an observer in isolation, this thesis incorporates higher-level EPS system requirements—including functional safety, production feasibility, computational constraints, and vehicle software–hardware integration—into the architectural decision process. These top-level requirements are systematically cascaded down to the estimator design through a structured top-down methodology, forming a key distinguishing feature of this study. By embedding cross-domain constraints into the SMO–HFI hybrid architecture, the proposed framework achieves more than signal-level estimation accuracy. It maintains coherence with implementation realities encountered in production-grade automotive systems, such as limited control-cycle budgets, inverter nonidealities, diagnostic requirements, and sensorless operational boundaries. This explicit traceability from system-level objectives to observer-level structure differentiates the work from conventional control-centric approaches and enables a balanced design that integrates academic rigor with industrial feasibility.