Social Media to Satellites

Currently, I am contributing to the ERC project So2Sat - 1016 Bytes from Social Media to Earth Observation Satellites as a senior scientist.

In this project, I am responsible for a subgroup working on social media text mining and social media image analysis as well as towards fusion of this information with remote sensing data.

I am especially working on the Big Geospatial Data aspects of this using the Linux Cluster and petascale compute platform SuperMUC at the Leibniz Supercomputing Centre.

Interdisciplinary Center for Applied Machine Learning (ICAML)

The Interdisciplinary Center for Applied Machine Learning aims at increasing the accessibility of machine learning across disciplines. Therefore, we conduct courses and events, collect and comment information, and provide a flexible platform for machine learning inside the browser.

The qualification within ICAML is envisioned to be as diverse as the various disciplines we want to connect. Many of our core qualification modules will be available in different formats: in extended formats for beginners or very concise (TL;DR) presentation.

The project works in tight collaboration with the Applied Machine Learning Academy (AMA).

We acknowledge the funding from the Federal Ministry of Education and Research.

A Selection of Completed Projects

Mobile Edge Computing (2013-2015)

In the development of mobile communication towards the LTE standards, the total bandwidth from base stations to mobiles is often by an order of magnitude higher than the bandwidth connecting the base stations with the core network. This and the generally high delay to the Internet limits the applicability of cloud computing in rural areas or for real-time applications. In this project, we implemented several applications that exploit the flexibility of the LTE mobile communication standard to provide services from a computing device directly attached to the eNodeB. We showed the feasibilty and usefulness of this approach with demonstration applications like content caching and video transcoding. Especially for video applications, the setup turned out to be very interesting as there is a tradeoff between computational complexity and the amount of data. In a mobile setting, computatinal complexity, however, translates to battery consumption. We showed that it is possible to use a very cheap encoding scheme on the mobile device leading to a high-bandwidth video stream which is transcoded on the base station into a highly compressed representation basically reducing both mobile energy consumption and effective bandwidth consumption in the core network.

Secure Graphics Card “What you see is what you sign!” (2012)

In this project, we implemented a complete state-of-the-art cryptographic system following RSA and PKCS.#1 which was able to run completely inside the NVIDIA CUDA compute architecture (>= 1.1). We not only demonstrated the possibility to do cryptography inside GPUs, we also implemented a prototype application that renders simple markup documents completely inside the GPU, awaits a PIN from the user, and signs the document. By integrating a secure element and a microcontroller managing a secure PIN keyboard into the GPU, this project envisiones to isolate the signature process from the (generally) unsafe computer platform. With hardware memory protection for the documents in the GPU and disconnecting the GPU from the PCI express bus during signature calculation, this would reduce attack surface from all components of the computer to the GPU and display subsystem.

Indoor Navigation at Munich Airport (2010 - 2012)

During my dissertation, we proposed and implemented the core of an indoor navigation system for Munich airport. The system was designed to remove the need for accurate indoor localization by following the principle of distributed display systems: the navigation interface is presented on displays that are distributed in space such that the users are able to find a display nearby (Ruppel, Gschwandtner, Schindhelm, & Linnhoff-Popien, 2009).

My part in this project was the design and implementation of the geometry backend for space-filling indoor navigation. The needed navigation data was generated in a semi-automated process starting from the CAD floorplans and building footprints. Special software modules were able to automatically detect functional objects such as buildings and staircases and the navigation system was able to cope with complex situations such that shops can be entered but not being used as a shortcut. In addition, visibility calculation was used in order to generate a set of landmarks. We also implemented a routing server which in addition to the shortest path and visibility calculation enabled the solution of partially ordered traveling salesman problems of varying sizes. These problems occur due to the user interface: in a video conference, a personal set of point of interests is produced including the wishes of the user (e.g., coffee, shopping, etc.) as well as points of interests related to the passenger process (e.g., checkin, dropoff, security check, etc.). The overall system design is context-aware such that delays and special rights (e.g., holding permission or keys) can be integrated at runtime.

The results of the project have been implemented by a Munich airport spin-off Infogate

  1. Ruppel, P., Gschwandtner, F., Schindhelm, C. K., & Linnhoff-Popien, C. (2009). Indoor navigation on distributed stationary display systems. In Computer Software and Applications Conference, 2009. COMPSAC’09. 33rd Annual IEEE International (Vol. 1, pp. 37–44). IEEE. [BibTeX]