Technology in our vehicle
In order to participate successfully in the university competition "Carolo-Cup" of the TU Braunschweig, the vehicle has to find its way independently in a street environment that is based on reality. The aim is not only to keep its lane at high speeds and follow a specified target trajectory, but also to overtake vehicles, detect intersections and park between boxes. All those actions that are carried out intuitively by a person, e. g. those that require a lane change, must be carefully analyzed, adapted and implemented. From obstacle detection to the actual lane change, many decision levels have to pass through. This includes all those components from setting the turn signals to determining a new target trajectory. The overall system comprises the three disciplines of mechanics, electronics and software.
The mechanical design of our system is of enormous importance. It has been shown that a robust basic structure is indispensable for every autonomous vehicle. Different requirements have to be worked out and implemented in the mechanical concept. For example, it has been shown that a shortened wheelbase has advantages when parking. A special feature of our team is the simultaneous use of front and rear axle steering. In order to take advantage of this advantage, components must be manufactured in-house. For example, a base plate for the chassis, battery holder, a camera and a bracket for fixing the encoder etc. are required. We also evaluate and implement unusual drive concepts such as individual wheel drives for a torque vectoring concept.
Various sensors are required for the perception of the environment and the state of the autonomous driving model car. For example, a rotary encoder for speed measurement as well as sensors for acceleration and angular acceleration measurement are used to keep the vehicle on track and the associated trajectory sequence control. For obstacle detection and parking, the vehicle requires distance sensors to the front, rear and sideways. The required sensors are read out with the aid of a microcontroller on a circuit board. In addition, the actuators, such as the motor for driving or steering the vehicle, are controlled and regulated by microcontrollers.
The software is very versatile and has to meet different requirements depending on the area of application. From hard real-time systems based on microcontrollers for the control of the actuators to particularly processor-intensive image recognition algorithms and machine learning solutions, we have it all. The camera is the primary sensor for the environmental perception of our vehicle. The camera detects lanes, obstacles, intersections and parking spaces. The image recognition data is fused with the other sensor data and an environment is modeled from it. An optimal and physically meaningful trajectory is then generated. The controller concept used generates the required manipulated variables from the trajectory, which in turn are passed on to the electronics.
We value particular importance to good software engineering. A modular structure of our software enables the independent development and exchange of individual modules. We also use common version control tools and an automated test pipeline.