Enhanced State Estimation Based on Particle Filter and Sensor Data With Non-Gaussian and Multimodal Noise

This paper presents a novel approach to state estimation based on particle filter dealing with measurement data effected by non-Gaussian, multimodal noise. The implementation focusses on autonomous underwater vehicles (AUVs) utilizing data of a magnetic compass and a mechanical scanning sonar for spatial navigation. Nowadays, particle filter approaches often require complicated feature extraction methods culminating in semantic interpretation of the data. This is not suitable for low-cost and low-weight AUVs, because these steps require high computational power. Therefore, efficient CPUs and higher power delivery are required. To test the novel approach, the algorithm is simulated in different scenarios with different parameters. Additionally, the filter is applied to real environment data. Finally, the performance is tested and evaluated by several methods. We demonstrate the computational efficiency and superiority of our method over other approaches through simulations.