a Improving the drivability of electric vehicles using advanced control and estimation

Abstract

The mechanical drivetrain dynamics of electric vehicles can have a detrimental effect on the driving experience and performance of the vehicle. Factors such as large gear backlashes, low damped axles and the relatively low inertia and friction of electric motors can lead to a drivetrain that is prone to oscillation and therefore giving a non-smooth acceleration response. When this is combined with unknown driving conditions, such as low traction surfaces and significant changes in the vehicle’s mass, this can cause the vehicle’s drivability to be poor, especially when closed loop speed control is required. Generally for all industrial vehicles and many on-road vehicles the only feedback to the vehicle motor controller is from the encoder on the motor, as it is already required for the flux vector control of induction and permanent magnet AC motors. Although the motor speed can easily be converted into a vehicle speed after taking into account gear ratios and tyre radius, it is not valid to assume that the motor and vehicle speeds are proportionally equal during transient conditions where there can be significant differences between them due to flexibility in the drivetrain. Electric vehicles offer improved efficiency over internal combustion engine (ICE) vehicles which can then be enhanced further through regenerative braking. Although this can lead to undesirable and dangerous conditions such as loss of traction during braking due to excessive regenerative braking, especially if the vehicle is rear wheel drive. Without any additional feedbacks it is traditionally assumed to be difficult to detect loss of traction or maximise the energy recovered through regenerative braking without risking wheel lock up. It has been shown that vehicle drivability can be greatly improved if estimates of vehicle speed and mass are obtained. Using a fixed gain Kalman Filter (KF) for state estimation and a Recursive Least Squares (RLS) scheme for parameter estimation, both vehicle speed and mass have been estimated and used to improve the closed loop speed performance. The estimated values are also used for preventing loss of tyre traction with the road surface. Additionally a scheme for reducing oscillations without estimating vehicle speed for use in torque control mode applications is developed. The Kalman Filter only works effectively when the correct process noise matrix Q and measurement noise matrix R are used, as well as the model being a reasonably accurate representation of the real system. Optimal tuning has been carried out using Genetic Algorithms (GA), with the estimation accuracy then analysed for robustness to varying vehicle mass. Gearbox backlash is the most significant issue within the drivetrain, as it is often the initial cause of the oscillations during torque reversals; allowing the motor speed to accelerate much faster than that of the vehicle during the disconnect. A scheme that can reduce the delay when traversing the backlash during torque reversals but also decrease the impact speed to virtually zero has been created.