Up to date, all assignments have been client specific. There is no standard service, but this page describes what Dreik Ingenjörskonst offers:

  • Programming in Octave and Matlab
    Most mathematical problems can be efficiently solved using Octave or Matlab, which are high-level languages for mathematical and generic programming problems. While being superior at short development time, performance may often be low compared to implementations made in C or C++. Paul has more than ten years of experience working with Matlab. He now primarily uses Octave,a free mostly Matlab compatible language. Octave can be extended by using so called oct-files, which is a very powerful method of extending octave with c++ code.
  • C++ development
    For performance critical projects, c++ is a good choice. See the c++ page for more information.
  • Machine learning
    Some problems can be efficiently solved by applying machine learning techniques. By learning from experience, a program can solve previously unseen similar tasks. See the track condition analyzer page for an example.
  • Probability analysis
    In many areas, uncertainty in measured or estimated values are important. By applying probability theory, consistent decisions can be made.
  • Running dynamics simulation
    Simulation of vehicles is a powerful tool. By simulation, properties can be studied with little need for expensive measurements and tests.
  • Equivalent conicity calculations
    See the equivalent conicity page.
  • Kalman filtering
    Being invented in the 60’s, the Kalman filter is still one of the most important methods for estimation and signal fusion. It is used to estimate unknown properties based on noisy measurements.
  • Particle filtering
    Some estimation problems are not possible to fit in to the model used by the Kalman filter. The particle filter is a general method to solve estimation problems, but is computationally expensive compared to the Kalman filter. Problems where the system evolves according to a nonlinear model and/or is subject to non-Gaussian noise are preferably approached with a particle filter.