INTRODUCTION Multi-axle land vehicles with drive actuation on more than one axle offer improved stability and traction on various road surfaces. This is possible via the optimization of load distribution and maximization of the utilization of individual tire-road contact by exploiting the redundancy of the drive system. The redundancy could also be exploited to boost fault tolerance and reliability. Going further, replacing traditional axle drive systems with wheel hub motors brings efficiency to the independent control of the wheel torque with fast and accurate actuation [ 1]. Nevertheless, the increased redundancy of independent hub motor control commonly results in an over-actuated vehicle system, which requires elaborate control allocation (CA) schemes. Ref [2] gave an overview of CA methods applied in vehicle motion control. The methods are often designed as part of a hierarchical vehicle control structure: at the top, a motion control algorithm is used to generate the global control efforts needed to satisfy the overall motion objectives; at the bottom, a CA algorithm coordinates the actuators to generate the desired global control efforts. Such a hierarchical design of CA system has been suggested in plenty of previous works concerning 4-wheel electric vehicles with access to torque control of hub motors [ 3,4,6]. This paper treats the case of multi-axle land vehicles with hub motor drives, which involve even more actuation redundancies. Researchers in [ 3] introduced a linear quadratic regulator (LQR) based dynamic yaw-moment control (DYC) method to generate global yaw moment needed to track the desired lateral response of a reference model under the input of driver’s steering and vehicle speed. Weighted least square method was used in [ 4,5] to resolve the distribution of longitudinal forces to each wheel from the global yaw moment under the hard constraint of the moment balance equation. The objective was set to minimize the magnitude of distributed force at each wheel, or slip ratio and its rates of change. Tire slip ratios, slip angles, and the states like vehicle slip angle and yaw rate are obtained by designing estimators [ 6,7]. In the above works, it was possible to obtain explicit solutions of the CA problem as linear tire model was used and the coupling between longitudinal and lateral tire force generation was not considered as a constraint. However, these simplifications are not practical in limit conditions outside of the regime of linearity. To ensuring the feasibility of the CA solution, global force tracking error can be added as a soft term in the objective function, as presented in [ 8], in addition to other terms might related to efficiency of tire usage [ 9] or energy dissipation of tire slip [ 10]. Optimization of this objective function was done subject to a linear approximation of friction ellipse constraint [ 11]. These resulted in a quadratic programming problem suitable for real-time application on a 4-wheel vehicle. The distributed forces were then used to generate the control variables based on a reversed tire model [ 9]. However, the linear approximation of the friction ellipse is conservative which leads to wasting part of the tires’ potential. Consider the nonlinear tire model as constraint in the CA can further improve the full usage of the tires, as pointed out in [ 12]. In this case, local controllers, like traction controller, could become of limited need as the wheel dynamics and its effects on the tire force generation are already integrated in the CA. The downside is the application of nonlinear tire model leads to Control Allocation for Multi-Axle Hub Motor Driven Land Vehicles Qian Wang and Beshah Ayalew Clemson University Amandeep Singh US Army, TARDEC ABSTRACT This paper outlines a real-time hierarchical control allocation algorithm for multi-axle land vehicles with independent hub motor wheel drives. At the top level, the driver’s input such as pedal position or s