GPS and LiDAR optimizing transformation parameters for localization in autonomous vehicles

Accurate localization is necessary for autonomous vehicles, with the demand for the correct fusion of Global Navigation Satellite System (GNSS) and Light Detection and Ranging (LiDAR) data. Existing static transformation parameter optimization methods do not work well to address dynamic environmental conditions such as GNSS signal weakening in urban canyons and LiDAR inconsistencies in open or obstructed environments. This work presents an LSTM-based technique of real-time transformation parameter optimization, automatically adjusting translation, rotation, and scale factors. The LSTM network processes sequential GNSS and LiDAR data, leveraging temporal correlations to enhance accuracy. Exhaustive experiments on real and simulated data demonstrate that the presented model reduces localization error by 25% compared to traditional techniques. The architecture provides an improvement of robustness over flexibility in complex situations like urban, rural, and tunneling conditions, and hence it is a strong solution for autonomous vehicle navigation