Thesis Work: CFAR RNN Radar Processor i Kungsbacka

Join ABB and work in a team that is dedicated to creating a future where innovative digital technologies allow greater access to cleaner energy.

ABB Jokab Safety R&D department develops functional safety products for use in industrial applications to avoid injuries and deaths. The products that are developed must be free from systematic faults, which means that the development must be state of the art, follow applicable standards and undergo rigorous third parity reviews and tests.

The ABB Jokab Safety product range is very wide, covering emergency stops to advanced safety PLCs. The products mainly consist of mechanics, HW electronics and embedded SW.

Driven by the availability of cost-effective high-performance sensors, small-scale radar systems have become ubiquitous in many applications, such as private cars and industrial automation. Nevertheless, the use of radar in safety systems has so far been limited. Unprecedented functionality may be realized with a multi-sensor radar system utilized for fast and accurate surveillance.

This thesis targets the reliability of a safety radar system, by proposing and simulating a local constant false alarm rate (CFAR) processor module for detection of protected objects (typically a human being), using recursive neural network (RNN) techniques.

Details:
• Period: Start at the earliest 2020-01-01, end at the latest 2020-12-31.
• Scope: 30 ECTS for Master Thesis.
• Number of students: 1-2
• Location: The most part of the work should be performed on site in Kungsbacka, 27 km south of Gothenburg (good public transportations).

• Review conventional CFAR techniques.
• Investigate whether CNN, in principle, is applicable for CFAR processing.
• Study benefits of a group radar CFAR processor.
• Propose a new, refined or modified CFAR algorithm, presumably based on RNN.
• Verify the performance of the proposed algorithm, contrasted to conventional CFAR, using a general purpose high-level mathematical software tool.
• Simulate various scenarios: Different filling ratio of surveillance volume with objects. Variable degree of interference, or signal-to-noise ratios.
• Author a report briefly describing state-of-the-art, discussing the rationale for the preferred CFAR processor in depth, assumptions and simulations of scenarios.
• Further define, extend and constrain the scope appropriately.

• Knowledge and some degree of fluency in statistics and signal processing.
• Demonstrated skills and interest in argumentation and documentation.
• Experience of and interest in high-level mathematical programming (Matlab/Octave/Python etc.) will facilitate the work.
• Ability to set up a rudimentary simulation environment where conventional and proposed CFAR algorithms can be benchmarked.
• Interest in with the closely related Thesis Work: Thesis Work: Safety Radar System Evaluation Unit. It will enable credible simulations but also understand inherent constraints and limitations of the targeted radar technology.
• Dedication to explore, describe and argue for mathematical algorithms.