Daniel Bräckelmann from Team ISF Löwen recently shared details of their vehicle used in the Carolo-Cup challenge earlier this month. Simba 2 won the hardest challenge of the competition, the Obstacle Circuit while placing third overall. Impressive to the Extreme. Looky here:
The Carolo-Cup is an annual international event for university students held by “Technische Universität Braunschweig” in Brunswick, Germany. The competition is inspired by the DARPA Urban Challenge and features three different challenges:
- Round Circuit – Drive as fast as possible on the circuit as possible for 3 minutes, while staying in the right lane of the course.
- Round Circuit with Obstacles – Same as the first challenge, but obstacles are distributed over the course. Some of the obstacles are in motion.
- Parallel Parking – Park as fast as possible in the smallest parking spot, parking in a large spot yields penalties.
In 2016, 15 student teams from all over Germany and a team from Gothenburg competed for two days in the Braunschweiger Stadthalle. The event is sponsored by many large German automotive manufacturers and part suppliers including Bosch, Infineon, Continental, BMW, VW and Daimler.
Löwen is German for “Lions”, which makes the name Simba, the lion cub from the film “Lion King”, a natural fit. The Simba 2 uses a stock Tamiya TRF417 for the chassis. Mounted to the chassis is a NVIDIA Jetson TK1 Development Kit, which is used for vision processing and algorithmic control of the car. Attached to the Jetson is a Teensy 3.2 µC for low level hardware interfacing with the steering servo, and the Electronic Speed Control (ESC) for the motor.
For the Carolo-Cup competition, a camera is mounted on a mast at the center of the car in addition to several other sensors (which appear to be IR and a laser line level). The camera is used for navigation, while the additional sensors are useful during the Parallel Parking part of the competition.
Team ISF Löwen has been working on the Simba project for around a year, with 10 students currently working on the project. You can sense the evolution of the project:
The software for the car was written from scratch using C++ and Qt. The resultant code is very lean (just about 5,000 lines of code).
A special ESC is used which allows very low speeds with high torque and fine-grain control. This is very useful in the Parallel Parking part of the competition:
For image processing we currently use OpenCV – CPU-only (Nvidias proprietary version – it’s just amazing how fast it is!).
Our plans for next years Carolo-Cup is to evaluate a complete GPU-based pipeline either using VisionWorks or the OpenCV GPU-module because we are close to the maximum performance the CPU can deliver (image size is 640×480 @ 50Hz).
There are also lights on the vehicle, which are used during the competition: