Localization Uncertainty Constrained Lateral PID Control with Aids of Fuzzy Logic Considering LiDAR NDT Matching Error

Localization and control are two key parts of autonomous driving. Accurate control relies on accurate positioning. Recently, the localization of autonomous vehicles based on the matching of Light Detection and Ranging (LiDAR) scan and High Definition (HD) map becomes the major solution. However, the matching can still possess meter-level positioning error in challenging areas with excessive dynamic vehicles or sparse features. Inaccurate positioning can result in obvious fluctuation in steering control of the vehicle subsequently, which is not acceptable for autonomous vehicles. In this paper, we propose to estimate the potential positioning uncertainty to further adaptively tune the parameters for the proportional-integral-derivative (PID) controller of vehicle steering. In this case, we can obtain a smoother control. Firstly, we generate the point cloud map of the tested area. Secondly, we correlate the uncertainty and optimal PID parameters using a fuzzy interference system. Finally, both the simulation and real experiments are conducted to validate the proposed method. The simulations show that the proposed adaptive PID controller is more resistant against unexpected positioning uncertainty and smoother control is obtained.