Cloud Intelligence and Collective Learning for Automated and Connected Driving

摘要

Moving a vehicle safely and automatically through road traffic is a complex task that even algorithms must first learn - with the help of various artificial intelligence methods. The huge volumes of data required for learning processes can only be processed by powerful cloud systems. In the future, fleets of automated vehicles will therefore be permanently connected to a cloud intelligence, as the ika describes it here. Artificial intelligence is a prerequisite for automatically solving any sufficiently complex technical task. Especially demanding are those tasks in which a set of multiple interacting problems has to be solved in a variety of environments. Automated driving is a textbook example for a complex task. Through the establishment of deep learning methods, which have raised the intelligence of technical systems to a new level, the goal of automating road traffic is now within reach. Artificial neural networks are capable of substan-tially increasing intelligence understood as the ability to solve problems in a variety of different environments [1]. Their high performance is primarily based on their ability to learn from large amounts of data. The rules for solving complex tasks are derived automatically, so developers no longer have to define these rules manually. Instead, they are now increasingly tasked with providing suitable data for training the neural networks. However, this shift in tasks is only a temporary step towards an even more comprehensive automation, which in the future will also include the continuous collection, preprocessing and provision of data for the training of models. In the future, the human development task will focus on the design of this automated process, which will continuously ensure and expand the performance of the learned algorithms, and make trained models available to the connected automated entities of the mobility system.