Optimizing machine learning algorithms applied to the context of intelligent vehicles

Computer Science & Engineering

Securing open networks, optimizing the performance of datacenters, and even getting more value from data are all part of the future of information technology. An important issue in its own right is the use and / or valuation of data produced and exchanged between these connected objects due to the number of nodes considered, the heterogeneity of the technologies for their connectivity, and the exponential evolution of needs expressed, given the volume of data available. Controlling and mastering connected and autonomous objects are becoming major societal and economic concerns. This has sparked intense activity in academia and industry in the area of machine learning research. Inside intelligent vehicle, machine learning algorithms are implemented in the detection, perception and decision subsystems to ensure localization, object recognition, trajectory planning, action prediction until obstacle avoidance. With the ultimate goal of optimizing these algorithms, we implemented cost functions, also called lost function, for a predictive model of a desired trajectory and motion planning. Moreover, several technics like dimensionality reduction, function approximation and model selection can be employed. We were able to determine the cost function of a cyberattack. Also, we used the dimensionality reduction technique for motion planning optimization to map the collision.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.