.png){kind=small}
Among all of Revolve NTNU’s technical divisions, the Control Systems group (CS) might just be the most difficult to understand. Whereas electrical and mechanical groups, such as Embedded Electronics and Suspension & Drivetrain, are directly responsible for building the things you can see and touch on the car, the influence of CS is not as visible. However, when the car is built and all systems are working, it is CS' responsibility to turn a fast car into a monster.
CS is generally responsible for the dynamic behavior of the car, i.e how the car moves when driving. Hardware was an important responsibility in the early days when the group designed parts of the suspension. The current group focuses predominantly on software with the core responsibilities being:
From these areas of responsibilities, one would be correct in assuming that CS is tightly connected to the field of cybernetics. Group roles change from year to year depending on the goals that are set during the planning phase, but a deep intuition and understanding of dynamic systems is a key trait that is developed by all group members. As a member of CS, you have the opportunity to not only enhance your engineering skills but also apply your theoretical knowledge to real-world challenges. For example, the laboratory assignments in the Linear Systems Theory and Optimal Control courses served as a strong influence in the development of Torque Vectoring. While the system models were different, the core ideas remain unchanged.
The effect of CS on the car becomes very visible once the car starts driving. While safety systems such as traction control and ABS are often taken for granted in modern consumer cars, CS develops these systems from the ground up. This allows us to build a lot more into these systems than you would find in a normal car. For example, our torque vectoring (TV) algorithm uses individual motor control to actively make the car turn and accelerate faster. Let’s take a closer look at TV and how it employs Linear Quadratic control.
TV is a program that allows us to vary the torque allocated on each wheel, where controlling the rotational motion around the car’s vertical axis is the main focus. To achieve this, a controller that compares the rotational requests from the driver with the angular velocity measurements has been made. We can increase this angular velocity by giving the outside wheels more torque and decrease it by giving more torque to the inside wheels. This angular velocity is called the yaw rate, and the “effort” to change the angular velocity is called the yaw moment, see Figure 1.
Control based on such a comparison is what we call feedback control, meaning that the difference between the requested yaw rate and the measured yaw rate will determine the yaw moment the controller outputs. A focus of this year has been to develop and test a variety of such controllers, where the LQR has been one of the concepts.
The LQR is a well-known control method used to regulate the behavior of a linear system. The objective of the LQR is to find the optimal control input, the yaw moment in our case, that will drive the system from its initial state to a desired state in the most efficient way. To do this, we define a cost function that measures the deviation of the desired state and the magnitude of the control input. The cost function is then minimized using the Riccati equation, resulting in a solution that provides the optimal control input. This means that the controller finds the optimal control input by balancing controller performance and the magnitude of the motor force.
To further improve performance, a reference feedforward was added. Feedforward control takes into account the future desired output, and uses it to pre-compensate the control signal. This results in a faster response to the system, which improves its overall performance. In the case of the car, adding a reference feedforward can ensure that the car converges to the desired reference in steady-state.
Feedback control is known for its accuracy, but can have slower response times. On the other hand, feedforward control is fast in response, but may not be as precise as feedback control. With the implementation of this controller, we hope to achieve a fast and accurate system that improves the car's ability to turn quickly. If done correctly, this will entail a decrease in lap times and improved overall car performance.
