Signal Processing Theory and Methods

Distributed learning algorithms based on methods such as least-mean square (LMS) or recursive least squares (RLS) were proposed not long ago, and allow sensor networks to collaborate to solve estimation and/or detection problems efficiently.  In addition, distributed versions of the Kalman filter were also proposed that allow optimal (or quasi-optimal) tracking of a parameter vector, under the constraints of linearity and Gaussianity.

Despite their superior performance, Kalman filters may not be adequate for use in sensor networks, due to their high computational cost  (the number of operations per sample needed grows with the cube of the number of parameters to be estimated), and also due to the need of a-priori knowledge of an accurate model for the parameter evolution.  We recently proved that it is possible to use combinations of adaptive filters to approximate the performance of Kalman filters with low computational cost (linear in the number of unknown parameters).  We also developed a new class of adaptive filters that allows near-optimal tracking with low computational cost for a larger class of models than possible with classical adaptive filters.

The goal of this work is to extend these results and to develop low-cost approximations to the Kalman filter that are robust against uncertainties in the model for the variation of the parameter vector to be estimated.  Due to the low complexity, this kind of algorithm would allow the implementation of algorithms with better tracking properties in sensor networks, and the robustness would allow the application to a class of problems in which performance guarantees are important, such as distributed control (cooperative autonomous systems, for example).

Candidates with a strong background in Signal Processing and Automatic Control or Telecommunications are encouraged to apply.  The position will be at the Electronic Systems Engineering Department at the University of São Paulo, São Paulo, Brazil, and may extend to up to four years.  The salary will be R$ 9,047.40 (tax-free), plus 10% for expenses.  Funding for relocation to São Paulo is also available, including for spouses.  

            Applications should be sent till Jan. 15, 2024 by e-mail to vitnasci@usp.br, with a short CV following the guidelines found here: 

https://fapesp.br/6351/instructions-for-the-elaboration-of-a-curricular-summary

The Signal Processing, Learning, and Computing (SPLC) Group at the Baltic Institute of Advanced Technology (BPTI) seeks a postdoctoral researcher in theoretical and applied signal processing.

Job Profile

The successful candidate will develop new computational methods for airborne object recognition in radar signals. At the technical level, the work will involve

• Performing theoretical analysis of relevant structural feature recoverability from Doppler signatures,

• Designing new specialized approaches to object recognition,

• Creating numerical algorithms for efficient implementations of object recognition on computing platforms,

• Carrying out real-world data analysis and model simulations.

There will also be possibilities to engage in other projects curated by the SPLC group or pursue your own ideas consistent with the group’s interests. Presenting research results in the form of scientific conference presentations and journal publications will be supported. Besides conducting scientific research, you will also have a chance to contribute to developing real-world systems with our industrial partners.

Requirements for the Candidate

• Doctoral degree in Applied Mathematics, Computer Science, Electronics & Electrical Engineering, Physics, or a related field

• Proven skills and experience in successfully designing and carrying out research projects

• Either theoretical knowledge or working experience in a significant subset of the following areas:

  • Numerical analysis,

  • Statistical inference,

  • Signal processing,

  • Mathematical optimization,

  • Machine learning,

  • Scientific computing.

• Proficiency in at least one high-level programming language and development environment suitable for numerical computing, data analysis, and visualization (e.g., MATLAB, Python, Julia) demonstrated via routine use in research projects

• Fluency in written and spoken English

Prior experience in radar signal processing is welcome but not required. A balanced experience in theoretical and computational topics is expected from the ideal candidate. However, people putting mathematical rigor in the first place are also very welcome to apply.

What We Offer

• 3500 – 4500 Eur / month (gross) salary depending on the actual competencies

• Possibility of qualifying for a permanent position after a 1-year fixed-term contract

• Focussing on quality research instead of chasing dubious performance metrics

• Close ties with industry to turn your research results into real-world solutions

Application

Please send the following information to open.positions.splc@bpti.eu compiled to a single pdf file:

• Cover letter briefly describing your motivation for choosing this position in connection to your interests and previous experience,

• Curriculum Vitae,

• A list of short summaries of your previous research projects describing the problem, the solution, the conclusion, and your contribution.

About Us

BPTI is a private, high-tech-oriented research institute established in Lithuania in 2008. We are driven by our mission to create value by providing R&D services for global security. Our core competencies in research and innovations lie in theoretical and applied machine learning and signal processing, integrated circuit design for radar and communication systems, cybersecurity solutions, and military UX/UI development. We believe that the origins of any technological breakthrough are rooted in fundamental research and, thus, dedicate a considerable part of our efforts to this dimension. We are a part of the official scientific research institution network of Lithuania.

The SPLC group at BPTI focuses on developing mathematical methods for extracting information from signals and other forms of data and implementing these methods numerically on computing platforms. In mathematical terms, our work revolves around solving various inverse problems of deterministic or stochastic nature. It builds on expertise in applied mathematics areas, such as numerical analysis, statistics, and mathematical optimization. We give equal emphasis to both the conceptual and the practical aspects of our work.