Moduł oferowany także w ramach programów studiów:
Informacje ogólne:
Nazwa:
Sztuczna inteligencja w pojazdach autonomicznych
Tok studiów:
2019/2020
Kod:
ZSDA-3-0005-s
Wydział:
Szkoła Doktorska AGH
Poziom studiów:
Studia III stopnia
Specjalność:
-
Kierunek:
Szkoła Doktorska AGH
Semestr:
0
Profil:
Ogólnoakademicki (A)
Język wykładowy:
Angielski
Forma studiów:
Stacjonarne
Prowadzący moduł:
dr hab. inż. Skruch Paweł (pawel.skruch@agh.edu.pl)
Dyscypliny:
Moduł multidyscyplinarny
Treści programowe zapewniające uzyskanie efektów uczenia się dla modułu zajęć

The aim of the subject is to address recent advances in Artificial Intelligence (AI) and Machine Learning (ML) that have potential to enhance the development of higher-level autonomous vehicles. It is anticipated that AI will enable vehicles to reach higher-level autonomous driving capabilities as AI based systems are more than a promising element to address technical challenges associated nowadays with autonomous vehicles.

Opis efektów uczenia się dla modułu zajęć
Kod MEU Student, który zaliczył moduł zajęć zna i rozumie/potrafi/jest gotów do Powiązania z KEU Sposób weryfikacji i oceny efektów uczenia się osiągniętych przez studenta w ramach poszczególnych form zajęć i dla całego modułu zajęć
Wiedza: zna i rozumie
M_W001 Student is able to properly formulate, describe and solve real problems connected with potential industrial applications SDA3A_W03, SDA3A_W07, SDA3A_W05 Kolokwium,
Prezentacja,
Udział w dyskusji,
Wykonanie ćwiczeń laboratoryjnych
M_W002 Student has knowledge how to design intelligent systems in the field of automotive engineering SDA3A_W07 Kolokwium,
Udział w dyskusji
M_W003 Student has knowledge on using Artificial Intelligence and Machine Learning based techniques in the development of autonomous systems SDA3A_W03, SDA3A_W07, SDA3A_W05 Kolokwium,
Prezentacja,
Wykonanie ćwiczeń laboratoryjnych,
Zaliczenie laboratorium
Umiejętności: potrafi
M_U001 Student is able to conduct applied research, studies and analysis and has the ability to combine research with development skills SDA3A_U07, SDA3A_U06, SDA3A_U01 Kolokwium,
Prezentacja,
Wykonanie ćwiczeń laboratoryjnych,
Zaliczenie laboratorium
Kompetencje społeczne: jest gotów do
M_K001 Student has the ability to work with people on efficient solving of research problems SDA3A_K01, SDA3A_K02 Prezentacja,
Wykonanie ćwiczeń laboratoryjnych,
Zaliczenie laboratorium
Liczba godzin zajęć w ramach poszczególnych form zajęć:
SUMA (godz.)
Wykład
Ćwicz. aud
Ćwicz. lab
Ćw. proj.
Konw.
Zaj. sem.
Zaj. prakt
Zaj. terenowe
Zaj. warsztatowe
Prace kontr. przejść.
Lektorat
30 15 0 0 15 0 0 0 0 0 0 0
Matryca kierunkowych efektów uczenia się w odniesieniu do form zajęć i sposobu zaliczenia, które pozwalają na ich uzyskanie
Kod MEU Student, który zaliczył moduł zajęć zna i rozumie/potrafi/jest gotów do Forma zajęć dydaktycznych
Wykład
Ćwicz. aud
Ćwicz. lab
Ćw. proj.
Konw.
Zaj. sem.
Zaj. prakt
Zaj. terenowe
Zaj. warsztatowe
Prace kontr. przejść.
Lektorat
Wiedza
M_W001 Student is able to properly formulate, describe and solve real problems connected with potential industrial applications + - - + - - - - - - -
M_W002 Student has knowledge how to design intelligent systems in the field of automotive engineering + - - + - - - - - - -
M_W003 Student has knowledge on using Artificial Intelligence and Machine Learning based techniques in the development of autonomous systems + - - + - - - - - - -
Umiejętności
M_U001 Student is able to conduct applied research, studies and analysis and has the ability to combine research with development skills + - - + - - - - - - -
Kompetencje społeczne
M_K001 Student has the ability to work with people on efficient solving of research problems - - - + - - - - - - -
Nakład pracy studenta (bilans punktów ECTS)
Forma aktywności studenta Obciążenie studenta
Sumaryczne obciążenie pracą studenta 90 godz
Punkty ECTS za moduł 3 ECTS
Udział w zajęciach dydaktycznych/praktyka 30 godz
Przygotowanie do zajęć 10 godz
przygotowanie projektu, prezentacji, pracy pisemnej, sprawozdania 15 godz
Samodzielne studiowanie tematyki zajęć 30 godz
Egzamin lub kolokwium zaliczeniowe 2 godz
Dodatkowe godziny kontaktowe 3 godz
Szczegółowe treści kształcenia w ramach poszczególnych form zajęć (szczegółowy program wykładów i pozostałych zajęć)
Wykład (15h):
  1. L01: Introduction and overview of autonomous vehicles

    First part of the lecture has the form of an introduction to the classes. The criteria for assessing and calculating partial and final grades are presented. Issues related to absence and improving of the classes are discussed as well. During this part a list of literature and other scientific aids is also provided.

    In second part of the lecture an overview of autonomous vehicles is provided. The goal is to describe essential elements of autonomous driving, such as: (1) sensors and perception; (2) computing platforms and control systems; (3) electrical architecture and network management; (4) vehicle connectivity; (5) user experience; (6) off board (cloud) support and services; (6) functional safety and security. In addition, technological challenges are highlighted where AI/ML based systems can play a crucial role. These challenges include (1) management and processing of humongous amount of data coming from different types of sensors in near-real-time; (2) development of a usable representation of the real-world operating environment to determine and predict further status of the vehicle; (3) making safety critical decisions in an uncertain operating environment; (4) building virtual simulators for test, validation and deep learning purposes; and (5) building a platform for the development of self-learning algorithms.

    Class duration: 4h

  2. L02: AI/ML based systems in autonomous vehicles

    The lecture is about elements (i.e., functions, features, components, etc.) of the automotive embedded systems that can be designed and developed with the help of AI/ML-based techniques in order to achieve higher performance comparing to traditional deterministic techniques.

    Class duration: 2h

  3. L03: Characteristics of datasets in automotive applications

    The main objective of the lecture is to characterize datasets that are currently used in automotive for verification and validation purposes and will be also used in near future for machine learning applications. The lecture provides a proposal how to formally, that is qualitatively and quantitatively by using proper mathematical notation, describe both static and dynamic nature of the data that cannot be omitted when designing algorithms.

    Class duration: 2h

  4. L04: Key performance indicators

    The lecture is about Key Performance Indicators (KPIs) that is about measurable values that are used in the automotive industry to evaluate perception algorithms of the interior and surrounding of the vehicle.

    Class duration: 2h

  5. L05: Meta learning and neural network topology optimization

    The lecture is about how to improve efficiency of the learning process and how to optimize neural network typologies taking into account time, size and performance constraints imposed on the system by the automotive industry requirements.

    Class duration: 2h

  6. L06: Explainable AI and interpretable ML

    The lecture is about AI and ML techniques that produce models that are more explainable and that enable human users to understand actions and decisions taken by these models. It is predicted that Explainable AI and Interpretable ML techniques will be essential in the applications for safety critical systems.

    Class duration: 2h

  7. L07: Summary and final test

    Summary of the lectures. Verification of learning outcomes by a final test.

    Class duration: 1h

Ćwiczenia projektowe (15h):
Design and implementation of neural network based classifier to detect a class of objects in the vehicle surrounding

The objective of the project classes is to solve the following research problem: for a given training data set design and train artificial neural network based classifier in order to detect a certain class of objects existing in the vehicle surrounding. The data set with labelled objects from the selected class will be provided. The neural network shall be designed and implemented in a dedicated machine learning framework that is based on Python and Tensorflow. The efficiency of the proposed architecture will be verified on a test data set using a set of KPIs that are embedded in the machine learning framework. The neural network architecture shall be optimized taking into account size, memory and performance constraints.

Pozostałe informacje
Metody i techniki kształcenia:
  • Wykład: Lecturing, demonstrating and collaborating
  • Ćwiczenia projektowe: Design thinking (case method)
Warunki i sposób zaliczenia poszczególnych form zajęć, w tym zasady zaliczeń poprawkowych, a także warunki dopuszczenia do egzaminu:

(1) Lecture part ends with the final test.
(2) Project classes end with the presentation on the resolution of the given research problem.
(3) The following criteria are taken into account when evaluating the presentation:
(a) values of the KPIs calculated on the trained neural network;
(b) size of the proposed neural network architecture;
(b) system resources (i.e., memory) needed to implement the solution;
© computation time.
(4) Every student has two opportunities to correct the final grade from the test and presentation. The grade after the corrections is calculated as arithmetic average of obtained grades.

Zasady udziału w zajęciach:
  • Wykład:
    – Obecność obowiązkowa: Nie
    – Zasady udziału w zajęciach: The attendance at the lectures is checked however participation in classes is not obligatory what means that the attendance will not affect the final grade. This rule does not apply to the last lecture with summary and final test.
  • Ćwiczenia projektowe:
    – Obecność obowiązkowa: Tak
    – Zasady udziału w zajęciach: Physical presence is not obligatory besides the last class. Information exchange, consultations and work progress monitoring are performed by means of electronic media. Student is obligatory to present his project results to other students on the last class.
Sposób obliczania oceny końcowej:

The final grade of the course is calculated on the basis of:
(a) results from the final test;
(b) results from the project classes.

The grade from the final test is a percentage value (min. 0%, max. 100%). The final grade from the project classes is also a percentage value (min. 0%, max. 100%).

The final grade of the course is calculated according to the formula:
final_grade [%] = 0.5*grade_from_the_exam_test + 0.5*final_grade_from_the_laboratory_exercises.

The final grade from the course as a percentage value (min. 0%, max. 100%) is translated to the following grading scale:
(1) from 91% bardzo dobry (5.0) [Polish equivalent of very good];
(2) from 81% plus dobry (4.5) [Polish equivalent of good plus];
(3) from 71% dobry (4.0) [Polish equivalent of good];
(4) from 61% plus dostateczny (3.5) [Polish equivalent of satisfactory plus];
(5) from 50% dostateczny (3.0) [Polish equivalent of satisfactory];
(6) under 50% niedostateczny (2.0) [Polish equivalent of unsatisfactory].

Sposób i tryb wyrównywania zaległości powstałych wskutek nieobecności studenta na zajęciach:

(1) If the student is not able to pass the test or provide the presentation on scheduled term for whatever reason, then another term shall be chosen upon agreement with the student, however in such case, the grade will be decreased by the value of 10%.
(2) Another date can be agreed only once.

Wymagania wstępne i dodatkowe, z uwzględnieniem sekwencyjności modułów :

(1) Bachelor or Master degree in control engineering, computer science or related.
(2) Good command of English.
(3) Familiarity with Python and Tensorflow.
(4) Familiarity with parallel and pipelined processor solutions.
(5) Knowledge of Artificial Intelligence and Machine Learning.

Zalecana literatura i pomoce naukowe:

1. Al Hakim, E.: “3d yolo: end-to-end 3d object detection using point clouds”. Master’s thesis, KTH, School of Electrical Engineering and Computer Science (EECS), Stockholm, Sweden, 2018.

2. Aptiv Autonomous Mobility (formerly nuTonomy): The nuScenes dataset, www.nuscenes.org, 2019.

3. Coursera: dedicated online courses related to data science, computer science, information technology, www.coursera.org, 2019.

4. Publications in scientific journals: (a) International Journal of Automotive Technology, (b) Journal of Vehicle Design, © Vehicle System Dynamics, (d) IEEE Intelligent Transportation Systems Magazine, (e) IEEE Intelligent Systems, (f) SAE Transactions.

5. Karlsruhe Institute of Technology: The KITTI Vision Benchmark Suite, www.cvlibs.net/datasets/kitti, 2019.

6. SAE International: “Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles”. Standard J3016 201806, www.sae.org/standards/content/j3016 201806, 2018.

7. Thakur, R.: “Scanning lidar in advanced driver assistance systems and beyond: building a road map for next-generation lidar technology”. IEEE Consumer Electronics Magazine, vol. 5, no. 3, pp. 48–54, 2016.

8. Tremblay, J., Prakash, A., Acuna, D., Brophy, M., Jampani, V., Anil, C., To, T., Cameracci, E., Boochoon, E., Birchfield, S.: “Training deep networks with synthetic data: bridging the reality gap by domain randomization”. Computer Vision and Pattern Recognition, arXiv:1804.06516, 2018.

9. Udacity: Nanodegree programs (Self-driving car engineer, Artificial intelligence, Deep reinforcement learning, Machine learning engineer, AI programming with Python, Deep learning, etc.), https://eu.udacity.com/nanodegree, 2019.

10. Vilalta, V., Drissi, Y.: “A perspective view and survey of meta-learning”. Artificial Intelligence Review, vol. 18, no. 2, pp. 77-95, 2002.

Publikacje naukowe osób prowadzących zajęcia związane z tematyką modułu:

1. Markiewicz, P., Dlugosz, M., Skruch, P.: “Review of tracking and object detection systems for advanced driver assistance and autonomous driving applications”. In: W. Mitkowski, J. Kacprzyk, K. Oprzędkiewicz, P. Skruch (Eds.), Trends in Advanced Intelligent Control, Optimization and Automation: Proceedings of KKA 2017 – The 19th Polish Control Conference, Kraków, Poland, June 18-21, 2017, Series: Advances in Intelligent Systems and Computing, vol. 577, pp. 224-237, Springer International Publishing, Switzerland, 2017.

2. Markiewicz, P., Kogut, K., Rozewicz, M., Skruch, P., Starosolski, R.: “Occupancy grid fusion prototyping using automotive virtual validation environment”. ICCMA Proceedings of the 6th International Conference on Control, Mechatronics and Automation, pp. 81-85, 12-14.10.2018, Tokyo, Japan.

3. Komorkiewicz, M., Turek, K., Skruch, P., Kryjak, T., Gorgon, M.: “FPGA-based hardware-in-the-loop environment using video injection concept for camera-based systems in automotive applications”. Proc. of the DASIP 2016 Conference on Design & Architectures for Signal & Image Processing, Book Series: Conference on Design and Architectures for Signal and Image Processing, pp. 183-190, 2016, 12-14.10.2016, Rennes, France.

4. Skruch, P.: “A complete deployment of model-based and real-time approaches in verification of production automotive embedded systems”. Proc. Of the 5th AutoTest Technical Conference on ‘Test of Hardware and Software in Automotive Development’, 15-16.10.2014, pp. 145-152, Stuttgart, Germany.

5. Skruch, P.: “Control systems in semi and fully automated cars”. In: W. Mitkowski, J. Kacprzyk, K. Oprzędkiewicz, P. Skruch (Eds.), “Trends in Advanced Intelligent Control, Optimization and Automation”, pp. 155-167, Springer, Switzerland, 2017.

6. Skruch, P., Buchala, G.: “Model-based real-time testing of embedded automotive systems”. SAE International Journal of Passenger Cars – Electronic and Electrical Systems, vol. 7, no. 2, 2014, doi: 10.4271/2014-01-0188.

7. Skruch, P., Dlugosz, R., Kogut, K., Markiewicz, P., Sasin, D., Rozewicz, M.: “The simulation strategy and its realization in the development process of active safety and advanced driver assistance systems”. SAE Technical Paper 2015-01-1401, 2015, doi:10.4271/2015-01-1401.

8. Skruch, P., Dlugosz, M., Markiewicz, P.: “A formal appraoch for the verification of control systems in autonomous driving applications”. In: W. Mitkowski, J. Kacprzyk, K. Oprzędkiewicz, P. Skruch (Eds.), “Trends in Advanced Intelligent Control, Optimization and Automation”, pp. 178-189, Springer, Switzerland, 2017.

9. Skruch, P., Dlugosz, M., Mitkowski, W.: “Stability analysis of a series of cars driving in adaptive cruise control mode”. In: W. Mitkowski, J. Kacprzyk, K. Oprzędkiewicz, P. Skruch (Eds.), Trends in Advanced Intelligent Control, Optimization and Automation: Proceedings of KKA 2017 – The 19th Polish Control Conference, Kraków, Poland, June 18-21, 2017, Series: Advances in Intelligent Systems and Computing, vol. 577, pp. 168-177, Springer International Publishing, Switzerland, 2017.

10. Skruch, P., Panek, M., Kowalczyk, B.: “Model-based testing in embedded automotive systems”. In: J. Zander-Nowicka, I. Schieferdecker, P. J. Mosterman (Eds.), “Model-Based Testing for Embedded Systems”, pp. 545-575, CRC Press, Taylor & Francis Group, Boca Raton, London, New York, 2011.

Informacje dodatkowe:

All documents related to the course will be placed on e-learning platform Moodle located under the following link: https://upel.agh.edu.pl/weaiib.