A Secure and Privacy-Preserving Collaborative Machine Learning System for Intelligent Transportation System

Autonomous vehicles assemble multiple intelligent mobility solutions to create an intelligent transportation system on the road. The intelligent mobility solutions across autonomous vehicles are achieved using advanced machine learning algorithms. The intelligence of the machine learning algorithms is proportional to the amount of variations in the training data fed into the algorithms. The static machine learning algorithms in the autonomous driving system fails to handle the new variations in the dynamic input data, which requires the continuous retraining of the machine learning models to adapt to the changes in input distribution. The dynamic driving environment through which the machine learning algorithms retrains differs from vehicle to vehicle. This difference in varied driving environments for different vehicles is reflected in terms of different retraining data for machine learning models which causes disproportional intelligence among difference vehicles for each driving functionalities. Here, the proposed system solves the intelligence mismatch problem by enhancing the intelligence of the machine learning models in different vehicles by transferring the learned knowledge from machine learning model of one vehicle to another across the distributed connected car system. The newly updated behaviors of all the driving functions are transferred across the cloud connectivity among the vehicles. The autonomous vehicles which are connected to one another and with the infrastructure via cloud help in the knowledge transfer from one vehicle to another. The autonomous driving functionalities such as perception and decision and control are improved by enhancing the intelligence of the machine learning models in vehicles through a knowledge transfer across the cloud which makes them omni-intelligent. The bidirectional knowledge transfer of the newly learned scenarios which are encountered during the driving through continuous retraining makes the global automotive transportation system efficient.