Artificial Intelligence in Automotive Technology

See TUMonline

After participation in the course, students will have a comprehensive overview of the methods of artificial intelligence and machine learning. Students are able to select the appropriate machine learning method for various problems and then implement it with the appropriate code. In addition, the students are able to tackle current problems in vehicle technology (e.g. autonomous driving) using machine learning methods.

The lecture covers all relevant aspects in the field of "Artificial Intelligence" with a special focus on "Machine Learning" and "Deep Learning" techniques. In addition, all theoretical aspacets will be related to automotive technology topics.

1. Introduction: What is Intelligence? What is artificial Intelligence? Historic overview, overview Machine Learning topics, self driving cars
2. Supervised Learning - Lineare Regression: Random Sampling & Consensus
3. Supervised Learning - Classification: Decision Trres, Support Vector Machines, k-nearest Neighbours.
4. Unsupervised Learning - Clustering: Decision Trees, k-Means
5. Path Finding: Navigation, Graph Theory, Search Algorithms like A*
6. Introduction to Neuronal Networs: Perceptron, Loss Function, Activation Function
7. Neuronal Networks: Backpropagation, Different Layers
8. Convolutional Neuronal Networks: Paramter, Filter, Visualization, Pooling
9. Recurrent Neuronal Networks
10. Transformers
11. Reeinforcemente Learning
13. AI-Development: Hyperparameter Tuning, Training on CPU and GPU, Inference

• Attendance of the lecture Basic of Motor Vehicle Contstruction
• Knowledge of programming with the programming language Python necessary and prerequisite for understanding the code examples performed in the lecture and exercise. We recommend an online course for Python e.g. at Codeacademy

In the lecture, the content of the course is taught by means of a lecture and presentation. More complex issues are derived and illustrated using tablet PCs. During the lecture questions are explicitly asked which expect a transfer payment from the students and which give the students the opportunity to speak and discuss a possible solution. The aim is to deepen the overview of the mechanical processes and to achieve the transfer for applying the mechanical processes to further problems. The lecture also explains simple code examples that can be actively programmed by the students. These code examples are primarily in the field of automotive engineering, which enables the students to work on special problems in the field of automotive engineering with machine learning methods. After each lecture unit, corresponding learning and programming tasks are handed over to the students in the form of a homework assignment, which deal with the subject matter of the learning unit and serve as preparation for the examination. For example, this is the detection of lanes in Chapter 2 Computer Vision or the detection of vehicles in Chapter 4 by Support Vector Machines. These programming tasks teach the students how machine learning methods can be converted into appropriate code and at the same time how to apply this to problems in vehicle technology.

In a written examination (duration 90 min) the taught contents are to be applied on the one hand to the basics of machine learning procedures as well as to various problems from vehicle technology and to be transferred to further tasks. For example, the students should prove in the exam that they have understood the basic mathematics behind the mechanical procedures and can apply them accordingly. The students should also be able to prove that they can select suitable machine learning methods for various problems in vehicle technology and implement them with the appropriate code. The calculator is allowed as an aid. By completing the homework after the lecture and submitting 50.00% correct results (calculated from the average of the percentage points achieved over all individual homework assignments), a grade bonus for the exam can be achieved.

Christopher M. Bishop Neural Networks for Pattern Recognition, 1995 /
Tom M. Mitchell, Machine Learning, 1997 /
Christopher M. Bishop, Pattern Recognition and Machine Learning, 2007
David Barber, Bayesian Reasoning and Machine Learning, 2012
Michael Nielsen Neural Networks and Deep Learning, 2014
Pendelten et. al, Perception, Planning, Control, and Coordination for Autonomous Vehicles, Machines 2017, 5(1), 6; https://doi.org/10.3390/machines5010006