On torque vectoring to improve steering predictability while minimising power loss in heavy electric vehicles using model predictive control

Abstract

The vehicle industry is moving towards electrical propulsion, and thereby improving the capabilities of the powertrain designs. Examples are the ability to produce both positive and negative torque, and the ability to have individually controlled wheels. These abilities also add redundancy to the powertrain system while allowing its primary objective of achieving the desired vehicle motion to be met. It may then fulfil secondary objectives, such as minimising power losses or tyre wear. With multiple motors, it is possible to micromanage the dynamics of the vehicle by varying the torque distribution on different wheels. This is called torque vectoring and includes the act of varying torque both in the longitudinal and lateral direction. This thesis analyses the impact of torque vectoring on a 4x4 heavy electric vehicle. The torque vectoring controller acts as a global force request generator, taking in a yaw rate reference together with either an acceleration request in form of pedal positions from the driver or a velocity request in an autonomous driving algorithm. The outputs of the controller are the requested global forces, which are then distributed to the actuators using control allocation, while optimising for power loss minimisation. A Model Predictive Controller (MPC), with a prediction horizon of 1 second, was chosen as the motion coordinator, as it is able to account for actuator and friction limitations and ensure that the requested forces to the control allocation are feasible. The target for the lateral control, in the MPC, was chosen as the vehicle’s steady state yaw rate response, as it reduces the amount of compensation from the driver during acceleration or braking. Results show that the lateral control from the MPC is able to reduce the braking distance in a curve, as well as reduce the steering wheel angle compensation from the driver when hard braking with 20-30 degrees. In addition, it is able to minimise the change in vehicle behaviour caused by changing distribution of propulsion between front and rear motors. In a simulation on road data from Hällered to Alingsås (distance of 2.8 km), the controller uses approximately 1.3% more power than a vehicle with no lateral control. An alternate controller style was developed to only interfere when the vehicle diverged from the expected behaviour. This controller managed to decrease the excess power consumption to 0.1% more than no lateral control.