Openings for postdocs

 

Information

Like for PhD students, almost all postdoc positions are linked to funded research projects. Positions become available once a project is funded. This can happen at any time during the year. Once a project is funded, the open positions will be filled as soon as possible. Typical appointment periods are for 1 to 1 1/2 years.

If you are interested in our research, it merits to inquire whether openings will be available. We collect resumes of prospective postdocs throughout the year.

Requirements

We make our selection based on the following requirements:

• Formal requirements regarding prior education: you should have earned a PhD thesis at a recognized university.
• Background: you should have a strong background in signal processing, communications or systems theory, and be proficient in linear algebra and mathematics.
• Excellence, as determined from your MSc Grade-Point-Average, awards and other recognitions.
• English: you should be able to communicate well in english (written and oral). Provide TOEFL/IELTS scores if available.
• Originality: your PhD thesis and related work (publications) should reflect original ideas that can be clearly pointed out. Critical and independent thinking is very important.
• Team player: you should be able to work well in a team of other project members.

Your resume should contain contact information of prior advisors/supervisors who can provide feedback.

Currently open positions

1. Dependable Distributed Sensor Systems [STW funded; project partners TNO, TU Delft Embedded Software group, UTwente Integrated Circuit Design group, TU Delft Circuits and Systems group]

   NATO Nationality Required

   Sensor networks have made a remarkable progress from a mere vision (Smart Dust, 1998), through early prototypes (Mica2 motes, 2002), to an ever-increasing set of deployments (2003 - today). However, operating a large-scale deployment involves still a lot of skilled manual labor, and experience is limited to static networks. This observation leads to the following, fundamental problem statement: How can we deploy, i.e. program and operate, large-scale distributed sensor systems across a wide variety of (unforeseen) circumstances such that users can depend on them.

   To keep this project tractable we propose an experimental approach around the application of localizing and tracking of people and objects in two complimentary scenarios: traffic monitoring and control in an urban environment (a fixed scenario with a dense number of sensors in a controllable environment) and rescue operations of firefighters and policemen (an ad-hoc scenario with few sensors in a hazardous environment). To realize working, and dependable, sensor systems in those scenarios we plan to contribute in three fields of research: the development of a miniaturized radar sensor, the development of an invironment to program large-scale distributed sensor systems, and the design and implementation of localization and tracking algorithms.

   This particular opening is related to the third topic, and more specifically to the question how different active sensing devices can cooperate in order to obtain an accurate position or track of the targets. To this end, we will focus on the two considered scenarios (traffic and safety). In the traffic scenario, the position of the sensors can usually be determined beforehand and the following problems will be tackled:

   • All the obtained range measurements have to be combined in order to obtain a reliable location or track of each target. This can for instance be done in a centralized fashion. In this case, all the range measurements are sent to a fusion center or sink (this can be one of the sensors), where the locations or tracks of the targets can be computed. We will investigate multi-dimensional scaling (MDS) algorithms and extensions thereof to solve this problem, as well as related dynamic versions of MDS for tracking purposes.
   • Using a fusion center or sink makes the system very sensitive to errors at this point. Hence, in this project we will also study decentralized localization and tracking. We plan to develop a decentralized algorithm that is similar in spirit to the decentralized algorithms for distributed estimation, where a global estimation problem can be solved in a completely distributed fashion using only local communications between neighboring sensors.
   • It is clear that not all sensors should participate in the localization or tracking process, only the ones that are close to the targets. Hence, we would like to investigate algorithms that can be used to activate only those sensors close to the targets in some adaptive mode. In other words, sensors should be turned on and off depending on the amount of information they can contribute.
   In the rescue scenario, the sensors are deployed only when needed, e.g., during a fire, and they should be able to exchange information in an ad-hoc fashion without a fixed infrastructure. As a result, to localize or track the targets, we first need to find out the coordinates of the sensors. The difference between localizing sensors and localizing targets, is that the sensors have a receiving capability and can actively cooperate with each other to determine their range to neighboring sensors. Once the ranges have been obtained, we can again use centralized or decentralized localization algorithms to determine the location of the sensors. Target localization and tracking will then proceed as in the first scenario.

   Status: Open

   Duration: Up to 3 years

   Contact: dr. Geert Leus

2. Distributed Signal Processing for Self-Organizing Wireless Networks [STW funded]

   A major problem in tomorrow's communications networks is the explosion of the number of devices. Preferably this connectivity is provided in a wireless fashion, but current wireless standards are not adequate to organize such large networks. The main reason is that they reserve a fixed amount of radio resources to the different devices, even if they are not in use, which results in a radio spectrum scarcity. Hence, the network gets saturated if too many nodes want to participate. Similar wireless connectivity problems appear in large sensor networks, where in addition information processing and localization become very problematic due to the large amount of sensors.

   The key to solve these problems is a paradigm shift towards self-organizing networks, where only local communication is allowed and each node adapts its procedures (related to spectrum utilization, sensing, information processing, and localization) based on information received from neighboring nodes and own observations. Instead of getting saturated by a large number of nodes, such networks benefit from a growing number of devices.

   To develop large self-organizing networks we need cognitive radio devices that are capable of sensing the radio spectrum and adapt accordingly. We further require energyefficient distributed information processing and localization algorithms for large sensor networks. The mathematical tools we want to build on are compressed sampling, convex optimization, game theory, and linear algebra.

   The proposed research will lead not only to new insights in self-organization, but also to novel algorithms and communications technology. It will result in a flexible, selforganizing, energy-efficient platform for a whole range of innovative applications, such as large-scale ad-hoc communications networks and plug-and-play wireless telemetry for monitoring and control, including tracking and tracing of devices.

   In this project, we have an opening for a postdoc with excellent track record in Signal Processing for Communications.

   Status: Open

   Contact: dr. Geert Leus

|  28 November 2011   |