1. INTRODUCTION Electric vehicles (EVs) have been of great interest during the recent decades, among which the ones driven by independently controlled in-wheel motors are considered as one of the most promising configurations for future vehicle design. With such configuration, the overall vehicle weight can be reduced, and there is more space for batteries, electric systems and auxiliary equipment, along with more cargo and passenger space. The merits of adopting electric motors as actuators are remarkable. By replacing the traditional engine-powertrain systems, the control torques, either accelerating or decelerating the vehicle, can be directly applied to every wheel. Thus, the power transmission efficiency can be highly improved, and the goal of zero-emission is ensured. Apart from all the advantages above, such configuration also provides an access for more accurate and flexible control to enhance the vehicle handling and stability performance. With the fast and precise response characteristics of the electric motor, unequal torques can be independently allocated, and adjusted according to the operation limits (e.g. road adhesion, actuator capability, etc.), or the control objectives (e.g. desired yaw motion, optimal path following, etc.). During the past decades, many optimal control method have been proposed and investigated to enhance the vehicle handling and stability performance based on tracking the desired yaw motion [ 1, 2, 3, 4]. In literature [ 5], an optimal controller was developed. The external yaw moment, which is to modify the vehicle handling and stability performance, consists of one feed-forward signal, the input steering angle, and two feed-back state variable, the yaw rate and the lateral velocity. In [6], an optimization-based yaw rate tracking algorithm was proposed, and implemented by generating unequal torques on the bilateral rear driven wheels via in-wheel motors. In [7], a PID-based yaw moment control system was designed, with its parameters optimized by genetic algorithm (GA). Comparing to general optimal control method, the PID controller is simpler to implement, meanwhile the robustness and the efficiency are ensured by the adoption of GA method for tuning the controller parameters. As an extension of traditional integer order PID controller, the Fractional Order PID (FO-PID) controllers have received much attention recently [ 8]. With more design parameters, such controllers possess flexibility and capability to fulfill more design specifications. However, this also raise the demand for an effective and efficient approach to automatically optimize the controller parameters. From the perspective of the design specifications of the closed-loop system, relay feedback tuning method is adopted in some literatures [ 9, 10, 11]. A more straightforward way is to directly use the optimization methods, e.g. Genetic Algorithms (EAs), Particle Swarm Optimization (PSO), etc., for searching the best set of the parameters. In [12], a new search heuristic named Differential Evolution (DE) is proposed. Comparing to traditional EAs and PSO, the DE algorithm has shown superior performance in some practical applications [ 13]. In this study, an auto-tuned FO-PID controller is designed to enhance the handling and stability performance of the EV prototype. The major contributions of this work are: 1. building an accurate multibody dynamics model with respect to the unique physical Design of an Adaptive FO-PID Controller for an In-Wheel-Motor Driven Electric Vehicle Yue Shi, Qingwei Liu, and Fan Yu Shanghai Jiao Tong University ABSTRACT An EV prototype, with all the wheels respectively driven by 4 inwheel motors, is developed, and undergoes a series of practical measurements and road tests. Based on the obtained vehicle parameters, a multi-body dynamics model is built by using SolidW orks and Adams/Car, and then validated by track test data. The virtual prototype is served as the control plant in simula