Autonomous Vehicles Trajectory Prediction Approach Using Machine Learning Test

Autonomous driving is the ability for a vehicle to sense the environment and act based on that without human involvement. Although this is useful and can provide many desired features to drivers, it still needs improvement and more investigation to provide safety since accidents could be fatal to human lives on top of the economical cost that resulted from it. Autonomous vehicles trajectory prediction could be used in reducing these losses by predicting the future path of the vehicle and the surrounding ones. In this work we addressed the problem of low accuracy machine learning trajectory prediction models, we propose an approach to implement a ML trajectory prediction model after applying Random Forest Regressor for feature selection and then using two ML classifiers (Decision Tree and Naive Bayes) to build the ML model. The best results were produced using the Decision Tree model without the feature selection phase (accuracy 99.99%, precision 100%, recall 99.99% and F1-score 94%). After using the feature selection, Decision Tree also scores the highest performance results (accuracy 94.80%, precision 78.30%, recall 80.10% and F1-score 79.20%) with execution time reduction 28.7% resulting of the Random Forest Regressor feature selection stage). After conducting the performance evaluation, it was shown that the proposed approach outperforms 3 previous approaches in the same research area.