127 - CROSS-DOMAIN NONLINEAR STATE ESTIMATION FOR AUTONOMOUS SYSTEMS USING UNSCENTED KALMAN FILTERING

About the role

This study focuses on developing a modular Unscented Kalman Filter (UKF)-based state estimation engine to enable robust, real-time navigation for autonomous systems in GNSS-challenged and sensor-degraded environments. The goal is to overcome limitations of traditional Extended Kalman Filter (EKF) techniques, which rely on local linearization and struggle with highly nonlinear or uncertain conditions, particularly when fusing heterogeneous sensors (e.g., GNSS, IMU, LiDAR, vision).

Responsibilities

• Define use cases and capture system requirements for UKF-based state estimation.
• Develop system models and sensor abstraction layers for multi-sensor fusion.
• Design and implement a modular UKF framework optimized for real-time performance.
• Deploy prototypes and conduct Hardware-in-the-Loop (HiL) testing.
• Validate the framework through field tests, including comparative benchmarking against EKF and other conventional filters.
• Assess impact and create a roadmap for commercialization or further development.

Key Innovations

• Nonlinear estimation without linearization, preserving model fidelity and accuracy.
• Adaptive noise modeling to enhance robustness in variable conditions.
• Modular, domain-specific architecture for integration into ADAS, autonomy stacks, or vehicle control systems.
• Platform validation, including optional demonstrations on ground vehicles and/or UAVs to illustrate domain portability.

Deliverables

• A validated UKF prototype for real-time multi-sensor fusion.
• Performance benchmarks comparing the UKF against EKF and other conventional filters.
• A reference dataset for testing in automotive (urban, rural, off-road) or aerial systems scenarios.
• An integration roadmap for commercial applications, including potential follow-up work on ISO-26262 compliance, automotive-grade hardware deployment, and open-source module release.