Ridge Regressive Data Preprocessed Quantum Deep Belief Neural Network for Effective Trajectory Planning in Autonomous Vehicles

Abstract

Trajectory planning is a new research topic in the field of automated vehicles (AVs). It is the process of identifying a trajectory for the vehicle to traverse its environment without obstacle collision. Trajectories are computed fast in real time as the environment constantly changes with time. To address these problems, the Ridge Regressive Data Preprocessed Quantum Deep Belief Neural Network (RRDPQDBNN) model is developed. The RRDPQDBNN model intends to carry out effective trajectory planning in autonomous vehicles through enhanced accuracy and minimum time complexity. Initially, in the RRDPQDBNN model, vehicle data are extracted and transmitted to the input layer. Secondly, Ridge Regressive Data Preprocessing is performed to eliminate noisy data from collected vehicle data. Finally, quantum data clustering is carried out in the RRDPQDBNN model to identify the severity of the risk without collision during the trajectory. This, in turn, is effective trajectory planning performed in autonomous vehicles. Experimental results are computed in terms of clustering accuracy, clustering time, error rate, precision, and recall. From experimental results, the RRDPQDBNN model increases clustering accuracy by 11%, precision by 13%, and recall by 5%, as well as reduces clustering time by 31% and error rate by 58% compared to existing methods.