Collision-Aware Autonomous Vehicle Motion Planning using Spiking Convolutional Neural Networks and Parrot Optimization Algorithm

The primary input for motion planning, which permits safe autonomous driving on public roads, is an accurate trajectory prediction of nearby road users. Advanced urban traffic conditions which are constantly evolving provide tremendous difficulties for autonomous vehicles specifically in the task of predicting the motion of surrounding vehicles and safely moving through it. This research seeks to address these challenges by presenting a new framework that combines a lightweight spiking convolutional neural network (LS-CNN) for real-time prediction of trajectories with an efficient collision avoidance technique based on the generated trajectory. The procedure includes learning from real-time information about traffic to estimate the direction of other vehicles around, and the parrot optimizer chooses the best routes of movement with low collision probability according to the results of LS-CNN perception. Computer simulations of the proposed motion planner in MATLAB software and in-the loop simulations show a significantly increased rate of collision avoidance and an efficiency of the paths generated by the proposed planner compared to conventional methods. The findings of the study emphasize that the integrated system improves both the trajectory prediction efficiency and the level of navigating safety. The model obtained above 95% of the trajectory prediction accuracy and also offered the path with the collision risk less than 0.005%. This work helps to provide an optimal performance for autonomous driving cars by implementing a highly adaptive traffic scenario prediction method to enhance the future enhancement of the autonomous vehicle navigation systems.