Automated drift control of an automobile
Winners of the competition
The winner of the competition is team YDrive from the Faculty of Electrical Engineering, University of Belgrade: Marko Skakun i Miloš Stojanović. Congratulations!
Results in the 2nd round
Faculty of Electrical Engineering, University of BelgradeMarko Skakun i Miloš Stojanović.
Marko Skakun i Miloš Stojanović.
KTH Royal Institute of Technology
|Ansh Gandhi, Ankur Fartyal, and Chitranjan Singh
Faculty of Electrical Engineering and Computing, University of Zagreb
|Vjekoslav Diklić and Marin Bogdan
Technical description summary
The goal of this challenge is to drive a given vehicle down a short, curved road segment, in such a way that the vehicle “drifts” as impressively as possible. To achieve this, contestants will need to design a control algorithm which intentionally drives the vehicle into an unstable equilibrium state and maintains this state through careful application of control inputs.
The algorithm will control a vehicle model, parametrized to reflect a drift car powered by electric motors. This model will feature realistic tyre models with lateral-longitudinal force coupling, vertical force dependency and saturation. The vehicle body and suspension will be simplified, to limit the scope of the challenge. The parameters of the vehicle model will be made available to all contestants, along with references to the equations the model was based on.
The vehicle is controlled by applying torque to the wheels and by rotating the steering wheel. Torque can be applied to any of the four wheels, in the positive or negative direction. The vehicle’s powertrain has a limited amount of torque it can supply per wheel, and the steering wheel can only be rotated up to a fixed maximum angle. Both torque and steering angle are not changed instantaneously – there is a maximum rate of change for both values.
The drift maneuver is judged objectively. The model computes and continuously outputs a cost function, whose value depends on:
- Δy – the lateral offset of the vehicle Center of Gravity from the road centreline
- β – the sideslip angle of the vehicle
- vx – the vehicle longitudinal speed
The challenge will consist of two stages. For Stage A, the contestants will be supplied with a vehicle model in the form of a Functional Mockup Unit (FMU) – a component based on the open FMI standard. This model can be integrated into a number of different simulation environments, including Matlab/Simulink. Contestants will be judged on the performance of their algorithm as applied to the provided vehicle model, according to the above described cost function. The top three teams will proceed to Stage B, and win a paid trip to Zadar to compete at the Conference.
Stage B will consist of implementing the algorithm in a format suitable for execution on a Rapid Development Platform (RDP), provided by an industry partner. The algorithm will be implemented in Simulink, and translated automatically to platform-compatible code using software tools provided by the industry partner. At the Conference proper, the finalists from Stage A will be provided with an FMU containing a differently parametrized vehicle model. In the time available, contestants will need to re-tune their algorithm and run it on the RDP, with the vehicle implemented on a Hardware In the Loop (HIL) test rig.
All resources related to the challenge are available in the following Github repository: https://github.com/rimac-technology/med2018-automotive-challenge
Rules of competition
The remainder of this page describes the specific rules for the Automotive challenge.
The challenge will consist of two stages. In Stage A, the performance of a control algorithm as applied to the vehicle model provided will be judged in accordance with the reward function. The top three (3) teams will proceed to Stage B, which will take place during the MED ’18 conference in Zadar.
Full technical description – Stage A
The document describing the Stage A (incl. vehicle model, track geometry, and reward function) is available at:
(Documents and software for the Stage B will be provided in due time to teams which quality for the Stage B.)
All resources related to the challenge are available in the following GitHub repository:
Dates and deadlines
Submission of simulation results: April 16, 2018
Stage A ranking announcement: April 30, 2018
Stage B (Finals): June 19, 2018