A Novel Machine Learning-Based ANFIS Calibrated RISS/GNSS Integration for Improved Navigation in Urban Environments

Area of Science:

• Robotics and Autonomous Systems
• Navigation and Positioning
• Machine Learning Applications

Background:

• Global Navigation Satellite Systems (GNSS) are crucial for autonomous vehicle (AV) navigation but suffer reliability issues in urban environments due to signal blockage and multipath interference.
• Traditional sensor integration methods for AVs often struggle to maintain accuracy under degraded GNSS conditions.
• Inertial Navigation Systems (INS), particularly Reduced Inertial Sensors Systems (RISS), offer an alternative but require precise calibration to mitigate drift.

Purpose of the Study:

• To propose and validate a novel sensor fusion approach for enhancing autonomous vehicle navigation accuracy and reliability in challenging urban environments.
• To introduce an Adaptive Neuro-Fuzzy Inference System (ANFIS) as a machine learning-based calibration technique for RISS.
• To evaluate the performance of the ANFIS-calibrated RISS/GNSS integrated system against traditional RISS/GNSS and Radar-based integrated systems.

Main Methods:

• Development of a navigation system integrating a Global Navigation Satellite System (GNSS) with a calibrated Reduced Inertial Sensors System (RISS).
• Implementation of a machine learning-based Adaptive Neuro-Fuzzy Inference System (ANFIS) for novel RISS calibration.
• Validation through real-world road trajectory tests and simulated GNSS outages of varying durations (50-150 seconds).

Main Results:

• The ANFIS-based RISS/GNSS integration demonstrated a significant reduction in 2D position Root Mean Square Error (RMSE) by 43.8% compared to traditional RISS/GNSS.
• A 28% improvement in 2D position RMSE was observed compared to a frequency modulated continuous wave (FMCW) Radar (Rad)/RISS/GNSS integrated system.
• The system achieved substantial reductions in 2D position maximum errors: 47.5% versus RISS/GNSS and 23.4% versus Rad/RISS/GNSS.

Conclusions:

• The proposed ANFIS-based RISS/GNSS integration offers superior positioning accuracy and reliability, crucial for safe autonomous vehicle operation in urban settings.
• The system exhibits long-term stability and is suitable for applications demanding continuous, precise positioning.
• The ANFIS calibration approach is extendable to other low-cost Inertial Measurement Units (IMUs), presenting a versatile and attractive solution for diverse navigation applications.