AbstractsEngineering

This dissertation addresses a thorough investigation and development on general problem formulations, control/optimization algorithm designs, and experimental validations of energy-efficient control allocation (EECA), with application to an over-actuated pure electric ground vehicle (EGV) with four independent in-wheel motors.For single-mode actuators which have only one working mode, either consuming or gaining energy, the existing CA scheme for over-actuated systems, which is usually in the form of different kinds of optimization problems, is extended to the EECA scheme by explicitly introducing efficiency functions of actuators as a portion of the power consumption expressions into the cost function. For dual-mode actuators which can work in both power-consuming and power-gaining modes, though not simultaneously, a virtual actuator concept and the corresponding compatibility condition is introduced to complete the novel EECA design frame in a general manner.Based on the proposed EECA scheme, different optimization algorithms with various characteristics are developed for the EGV in order to achieve real-time implementations. For the vehicle longitudinal motion control, since the front and rear two in-wheel motors of the EGV can be lumped together as two pairs, a Karush-Kuhn-Tucker (KKT)-based EECA algorithm is proposed to find all the local optimal solutions, and consequently the global minimum through a further simple comparison among all the realistic local minima and boundary values for a non-convex optimization problem. The KKT-based algorithm is also independent on the selections of initial conditions by transferring the standard nonlinear optimization problem into classical eigenvalue problems. Consequently, the KKT-based algorithm is real-time implementable and is validated through both Simulink-CarSim® co-simulations and experimental results. Since the computational load will exponentially grow along with the increase of the number of actuators, the KKT-based algorithm is not suitable for the planar motion control of the EGV, in which four independent in-wheel motors are applied. Thus, an adaptive EECA, which introduces a proper Lagrangian function into the Lyapunov approach, is designed to tackle the real-time implementation challenge for the EGV planar motion control. In the adaptive EECA, the control distribution can be dictated to gradually converge to the energy-optimal operating points without solving the complicated nonlinear or non-convex optimization problem at every instantaneous sample time. Moreover, taking allocation errors between virtual control efforts and real actuation realizations as inputs, the input-to-state stability of the overall feedback system is proved.Simulation and experimental results from a four-wheel independently-actuated in-wheel motor electric vehicle also validated the effectiveness of the abovementioned EECA method.