Study of wheel slip and traction forces in differential drive robots and slip avoidance control strategy* Edison Orlando Cobos Torres, Shyamprasad Konduri, and Prabhakar R. Pagilla 1 Abstract— The effect of wheel slip in differential drive robots is investigated in this paper. We consider differential drive robots with two driven wheels and ball-type caster wheels that are used to provide balance and support to the mobile robot. The limiting values of traction forces for slip and no slip conditions are dependent on wheel- ground kinetic and static friction coefficients. The traction forces are used to determine the fraction of input torque that provides robot motion and this is used to calculate the actual position of the robot under slip conditions. The traction forces under no slip conditions are used to determine the limiting value of the wheel torque above which the wheel slips. This limiting torque value is used to set a saturation limit for the input torque to avoid slip. Simulations are conducted to evaluate the behavior of the robot during slip and no slip conditions. Experiments are conducted under similar slip and no slip conditions using a custom built differential drive mobile robot with one caster wheel to validate the simulations. Experiments are also conducted with the torque limiting strategy. Results from model simulations and experiments are presented and discussed. I. INTRODUCTION Research related to mobile robots in a variety of areas, such as dynamic modeling, control design, coordination, has been active in the last several decades. Because of their simplicity in construction and dynamics, differen- tial drive mobile robots have been the most common configurations considered. A typical differential drive robot consists of two driven wheels and one or more caster wheels. Most of the research in differential drive robots has assumed pure rolling of robot wheels, that is, all the torque provided to the robot wheels is used to provide motion to the robot. Kinematics and dy- namics of these types of mobile robots and controller designs for achieving various motion objectives have been considered in literature [1]–[4]. In practice the pure rolling assumption does not always hold, especially in circumstances where large values of input torque are used to accelerate the wheels, thereby resulting in wheel *This work is supported by National Science Foundation under Grant No. 0825937. 1 E. O. Cobos, S. Konduri and P. R. Pagilla are with the School of Mechanical and Aerospace Engineering, Oklahoma State University, Stillwater, OK 74078, USA pagilla at okstate.edu slip. The inability to reach the desired position due to wheel slip has been reported in the literature; such examples are presented in [5], [6] where wheel slip causes problems in tracking and the inability to main- tain desired distance between vehicles in coordination maneuvers. Wheel slip can occur in the longitudinal and/or lateral direction of the wheel motion. It depends on the wheel accelerations and the traction forces between the wheels and the ground. The interaction between the ground and the wheel could be complex depending on the various types of wheel and ground properties that are consid- ered; for example soft rubber covered wheels, loose soil, etc. There are several theories and approximations to describe these interactions [7]; many studies use the properties of tire and ground to determine the coefficient of friction and traction forces. Although these models provide considerable understanding of slip, they require a number of parameters to characterize the wheel and ground behavior during motion. Methods have been developed for predicting and controlling the behavior of the robot when it slips [8]–[14]. In these studies the kinematics and dynamics under slip conditions are modeled by considering either a flexible wheel, rigid ground or rigid wheel, flexible ground interaction. Some literature in the last decade has focused on rigid wheel, rigid ground interaction [5] and [15]. In [5] a relationship between the traction forces and applied wheel torques is obtained without considering the portion of mass supported by each wheel; a two stage controller is employed for trajectory tracking. It is indicated that the measurement of the global location of the robot is necessary to effectively control the robot. In [15], the traction forces are determined by using the static friction coefficient between the wheel and the ground for both slip and non-slip conditions. In previous studies, the effect of the caster wheel is ignored when calculating the normal forces. In this work, wheel slip with a rigid wheel, rigid ground interaction is considered, since it best represents the wheel-ground interaction of the robots used as part of this research. The traction forces are determined by 2014 American Control Conference (ACC) June 4-6, 2014. Portland, Oregon, USA 978-1-4799-3274-0/$31.00 ©2014 AACC 3231