Heading Angle Estimation Using Intelligent Fusion for Autonomous Vehicles

Low-cost navigation systems for ground vehicles often rely on the fusion with Global Positioning System (GPS) for improved state estimation. In this study a low-cost gyroscope is fused with a GPS for improved vehicle heading angle estimation using several Artificial Intelligence (AI) based architectures that include Shallow Neural Networks (SNN), Multi-Layer Neural Networks (MLNN), and Adaptive Neuro- Fuzzy Inference Systems (ANFIS). The primary goal behind using AI based sensor fusion is to obtain a highly accurate vehicle heading estimation suitable for autonomous navigation applications. When available, the GPS signal is used to correct the vehicle's heading angle. The neural networks and ANFIS methods both use the difference between the GPS signal and the heading angle from the integrated gyroscope as inputs. The performance achieved is shown and analyzed. According to the results obtained, the MLNN provides the most accurate heading estimates.