The TUfast Eco Team is a student team with around 40 members who design and build a highly efficient, electric car every year. In 2016, the team was able to set the world record for the most efficient electric car with a reach of 1232,27 km/kWh (which equals 10957,02 km/l Super 95) with the prototype "eLi14". In 2017, the team built their first UrbanConcept car called the "muc017". This category aims for the concept of a car which could one day drive around in our cities. This means that, amongst others, the car needs to have space for luggage, lightning, and an upright seating position.
For the 2018 Shell Eco-marathon Europe, the TUfast Eco team has built a new UrbanConcept, called the "muc018", with autonomous driving capabilities. The competition consists of various challenges, such as a track drive on the efficiency circuit. For all challenges, a precise estimation of the state of the vehicle is essential. For the 2018 season, the TUfast Eco Team is supported by the Xsens MTi-G-710-GNSS IMU, which is mounted in the center of the car.
Estimation of vehicle parameters
One of the main challenges before the competition is precisely estimating the vehicle parameters. The MTi-G-710-GNSS IMU gives extra information, in addition to odometry, we need to calculate the values.
Vehicle state for autonomous driving
To implement self-driving functions, a precise estimation of the vehicle state is indispensible. This includes not only the acceleration and yaw rate, but also the absolute position and orientation of the vehicle. This is all realised using a self-developed Kalman filter, which fuses the IMU data with the vehicle’s odometry. The obtained vehicle state is then used for mapping and for the control system.
Validation of the state estimation
On a test-track, a rectangle of approx. 100m length, the approach has shown a deviation of less than 50cm. Also on-track, the system has shown a good performance both during both the trackdrive and the "parking challenge". The trackdrive features a long distance and an angle difference of 360° while the "parking challenge" requires precise estimates for close manoeuvering. We could not have done this without the MTi-G-710-GNSS. Especially the tight coupling with the GNSS receiver, allowing to estimate the yaw angle precisely over a long distance, proved to be very important!