Soft Constrained Autonomous Vehicle Navigation using Gaussian Processes and Instance Segmentation
This work addresses localization challenges for autonomous vehicles, but it is incremental as it builds on existing feature-based navigation methods.
The paper tackled autonomous vehicle localization by using instance segmentation and Gaussian Process Regression to predict vehicle-to-landmark distances, resulting in improved prediction accuracy and reliable localization with fewer landmarks and noisy observations.
This paper presents a generic feature-based navigation framework for autonomous vehicles using a soft constrained Particle Filter. Selected map features, such as road and landmark locations, and vehicle states are used for designing soft constraints. After obtaining features of mapped landmarks in instance-based segmented images acquired from a monocular camera, vehicle-to-landmark distances are predicted using Gaussian Process Regression (GPR) models in a mixture of experts approach. Both mean and variance outputs of GPR models are used for implementing adaptive constraints. Experimental results confirm that the use of image segmentation features improves the vehicle-to-landmark distance prediction notably, and that the proposed soft constrained approach reliably localizes the vehicle even with reduced number of landmarks and noisy observations.