TUD

Institute of Automation

Modellbasierte Systemanalyse

Studien- und Diplomarbeit

MOBATSim is a simulation framework based on MATLAB Simulink that allows the user to assess vehicle level and traffic level safety by a 3D traffic simulation. It is used to simulate urban city traffic, design intersection management algorithms for the infrastructure or path planning algorithms for vehicles and test the efficiency of autonomous driving functions by PC-based simulations. Simulink 3D Animation and V-Realm are used to visualize the driving scenarios.

MOBATSim is still being developed to make it a complete tool for comprehensive testing. Its main task is to test autonomous vehicles and their functions in the presence of various faults. New fault libraries are being defined. An automatic report generator is being developed in accordance with the ISO 26262 standard. Possible Studien- and Diplomarbeit topics can be given to the students who are interested in autonomous vehicle modeling, simulation, and safety assessment.

The students who want to be a part of the project should know coding and modeling in MATLAB and Simulink, and be highly proficient in both spoken and written English. Nevertheless, the Studienarbeit or Diplomarbeit can be written in German. Detailed information about MOBATSim can be obtained from: https://mobatsim.com.

To watch the award winning video of MOBATSim for the worldwide Simulink Competition 2018:
https://www.youtube.com/watch?v=rG8B0ip4dpk

Contact: Mustafa Saraoğlu, M.Sc.
Mustafa.saraoglu (at) tu-dresden.de

Studien- und Diplomarbeit

SafeTown project shows how Lego EV3 Robots, UDP and a detector camera can be used to simulate a SafeTown using Simulink models. SafeTown is a project where we only use Simulink and Add-on Toolboxes to control a group of autonomous vehicle robots (LEGO EV3) and manage the traffic on a small scale map using image recognition with a camera and a workstation. We use all of them in coordination using real-time communication (UDP). The challenge in this project is to apply the developed models on Simulink to real life where the hardware components cause a lot of unexpected troubles, such as noisy sensors, changing environmental conditions, and the effect of the changing states of batteries on the vehicles. The goal is to have a "SafeTown" where autonomous vehicle robots drive freely without colliding with each other. This project is open to innovations from multidisciplinary fields such as: sensor fusion, computer vision, deep learning and reinforcement learning

The traffic should be controlled using a camera mounted on top, trying to recognize vehicles and send pass/wait commands through a workstation to the vehicles using UDP. The students who want to be a part of the project should know coding and modeling in MATLAB and Simulink, and be highly proficient in both spoken and written English. Nevertheless, the Studienarbeit or Diplomarbeit can be written in German.

To watch the award winning video of SafeTown for the worldwide Simulink Competition 2019:
https://www.youtube.com/watch?v=Vsq4WfuFVyE

Contact: Mustafa Saraoğlu, M.Sc.
Mustafa.saraoglu (at) tu-dresden.de

Efficient Testing of Automotive Model-based Software

Up to 80% of the automotive software can be generated from models. MATLAB Simulink is a common tool for creation of complex combinations of block diagrams and state machines, automated generation of executable code, and its deployment on a target ECU. The automotive safety standard ISO26262 requires extensive testing of the developed models with a large number of test cases. These activities can account up to two-thirds of the cost of software production. Automated software tools like Simulink V&V, Reactis Tester, TraceTronic ECU-TEST or TPT allow the generation of a complete test suite with maximum coverage and minimum redundancy. However, the testing stays extremely time-consuming even these tools. The topics, listed below, are devoted to the improvement of the model-based automotive testing with several intelligent methods. The tasks are closely connected to the industry. Students will get an opportunity to work with real automotive models and tools provided by our partners from TraceTronic and BMW.

Topic 1: Automatic Fault Localization

In practice, the testing tools basically show which test cases were successfully passed and which were failed. We are developing a new method that allows the identification of the faulty part of the model, based on the structural and behavioural properties of the model and the information about the succeeded and failed test cases. The intended method will be based on recently introduced stochastic dual-graph error propagation model and incorporate intelligent methods such as model reduction, forward propagation of data diversity, and backward error propagation analysis.

The task is to implement a prototype of the fault localisation method (or a part of it) and demonstrate it with a realistic Simulink model accompanied with a number of test cases provided by our industrial partners.

The student will work with the following methods and technologies: MATLAB, Simulink, Simulink API, Python, Basic probability theory, Markov chains, PRISM model checker and testing tools such as TraceTronic ECU-TEST and Reactis Tester.

Contact:
Dr.-Ing. A. Morozov

Topic 2: Test-case Prioritization for Regression Testing

Regression testing, as it is shown in the figure above, should be undertaken every time the models are updated to ensure that the modifications do not introduce new bugs into a previously validated model. A common, time-consuming way is to rerun an entire test suite after even minor changes. We develop a new method for automatic prioritization of test cases for efficient regression testing. The method is based on a recently introduced dual-graph error propagation model - a stochastic mathematical framework that describes data error propagation processes. The method automatically estimates which parts of the system can be affected by the errors that have occurred in the updated blocks and identifies the test cases that will both stimulate errors in the updated components and detect the occurred errors.

The task is to optimize and extend the current implementation of the prioritization method towards full-scale automotive models. Our colleagues from the TraceTronic Dresden will provide realistic Simulink models used by BMW and ensure that the method is implemented according to industrial needs.

The student will work with the following technologies and methods: MATLAB, Simulink, Simulink API, Python, Basic probability theory, automated testing tools.

Contact:
Dr.-Ing. A. Morozov

Last modified: 28.01.2020 14:38
Author: Webmaster IFA