Navigation

Minhao Qiu, M. Sc.

Department of Computer Science
Chair of Computer Science 7 (Computer Networks and Communication Systems)

Room: Room 06.125
Martensstr. 3
91058 Erlangen

More Information

No publications found.

  • Reliability design of multi-sensor systems

    (Third Party Funds Single)

    Term: 2020-01-01 - 2022-12-31
    Funding source: andere Förderorganisation
    Modern driver assistance systems for self-driving cars often rely on data collected by different sensors to determine the necessary system decisions. To prevent system failures, different techniques can be used to enhance the reliability of such multi-sensor systems, e.g., aggregation, filtering, majority voting and other mechanisms for fault tolerance. As a consequence, erroneous sensing is rare but can be correlated in successive sensor readings (e.g., as error bursts) and also between sensors (e.g., in specific environmental conditions such as bad weather).

    For a reliable design, error probabilities of such multi-sensor systems must be determined. In the project an existing analytical model based on Markov chain as well as a simulation model should be extended. This includes the following aspects: extensions for several correlated sensors, integration of practically relevant sensor fusion algorithms, consideration of environmental conditions, adaption of structure and parametrization of the error model, extensions of the simulation for rare events and inclusion of code, validation of the model results based on available data, and realization as a software tool for the reliability design of multi-sensor systems.

    In this project, the preliminary work of the INI.FAU project is to be built and both the existing analytical model based on Markov chains and the simulation model for multi-sensor systems are to be expanded. The desired scientific knowledge consists in the further development of the analytical Markov model, which already takes into account bursts of errors of individual sensors and dependencies between two sensors, the expansion to more sensors, the consideration of further error prevention strategies and a tool implementation. Furthermore, knowledge of the use of rare event simulation is to be achieved in order to execute more detailed simulation models of multi-sensor systems in practical terms and thus to derive statistically reliable results. The simulation allows an even more realistic system simulation and a validation of the analytical modeling. A scientifically based methodology is developed to determine the reliability of multi-sensor systems.