Hello,
i’m Gregor Godbersen, a full-stack developer, entrepreneur, consultant and research associate at TUM. On this seldom updated private blog you will find some thoughts and neat projects. Feel free to contact me.
Articles
-
This year I participated in the annual Kaggle Christmas Challenge and will illustrate my approaches in this post since the submission deadline has passed. My final model reaches the optimal solution within 37 hours using a linear programming min-cost-flow formulation. A 3% optimality gap is reached within 5 hours.
Read more -
The tiny and cheap ESP8266 WiFi-enabled microchips allow for quick hardware prototypes. Among the libraries that have sprung up in the considerable ecosystem is the graphic library minigrafx that supports several LCD, OLED, and E-Paper displays. The API provides several graphics primitives as writing text, drawing lines and rectangles, which can be composed into complex user interfaces.
Read more -
This project aimed to create a fully automated pipeline running on a RaspberryPi to digitalize incoming mail, documents, and receipts at the push of a button. The code can be found at github.
Read more -
While several bike-sharing operators have been active in Munich for some time (Callabike launched in 2000), the concept has recently returned to the spotlight through the new market entrant oBike. Having launched in August 2017, it quickly flooded the streets with its signature yellow bikes. Early in the launch phase, it was unclear just how many bikes were deployed, with low numbers around 350 being communicated and city officials being unsure about further plans (SZ.de)
Read more -
We were working with the Xilinx zturn7 FPGA development board from MYiR, which was running an old linux kernel for a start-up project. To update the board and build an custom embedded linux distribution using Yocto Project, we needed to know which modifications from stock Linux were made.
Read more -
This post is based on a recent problem I encountered while working on an academic question. We had a frequently changing set of input cases, each describing a specific transportation problem’s layout. The cases were sent to an algorithm chain and discussed within the technical documentation. To prevent stale data from entering the report, we wanted to create a solution that would use the input files used for the algorithm to be the source for the report, thus having a single truth source. The goal was to replicate the following illustration using only a CSV file containing the coordinates and types of the nodes using LaTeX and PGF/TikZ. (LaTeX is a wordprocessor and markup language that focuses strongly on separating content and layout while TikZ allows for the creation of vector graphics within LaTex using a descriptive language.)
Read more -
I wanted to visualize the paths taken by riders of the Munich bike sharing system. Unique about this system is that there are no fixed stations as bikes can be rented and returned at any street corner within the city limits, resulting in much more interesting data than in station-based systems. As the bikes register their coordinates when free, repeated polling of the free bikes on the system’s website and tracking their locations can give reasonably accurate measures about the trips taken. I used openstreetmaps cycle-path data to turn the start and end locations into approximations of the routes taken. While this allows for several statistics to be made about the driving habits, the most visual result is a live replay using the recovered times and routes, which I have embedded below. Note: This is a write up of a project I did in 2014 so the data is a bit older
Read more -
While standard PDF annotations are well supported in PDF clients and browser plugins, they do not allow for rich annotations in a web context. For this project, I explore the possibilities of an online presentation annotator. The basic idea was to allow users to mark sections within each slide and to attach a written note to it. For the PDF rendering, two different modes were tried, one that pre-renders the pages as images and provides meta information through a JSON file, the other using Mozilla’s PDF.js to render the PDF pages directly in the client.
Read more
subscribe via RSS