Robotica (2006) volume 24, pp. 333–335. © 2005 Cambridge University Press doi:10.1017/S0263574705002225 Printed in the United Kingdom Trajectory optimization of flexible mobile manipulators H. Ghariblu and M. H. Korayem (Received in Final Form: August 7, 2005, first published online 31 October 2005) SUMMARY A computational algorithm is developed to find a dynamic motion trajectory of a mobile manipulator with flexible links and joints that will allow the robot to carry a maximum load between two specified end positions. A compact form of the linearized state space dynamic equations is organized as well as constraint equations. Then, the problem of finding a maximum load carrying capacity on flexible mobile manipulators is formulated as a trajectory optimization problem. KEYWORDS: Trajectory optimization; Mobile manipula- tors; Linearized trajectory. I. INTRODUCTION Finding the full load motion for a specified point-to-point task can increase the productivity and economic usage of mobile manipulators. In a classical fixed base manipulator, the dynamic load carrying capacity (DLCC) is defined as a maximum load, which a manipulator can carry repeatedly in its fully extended configuration, while the dynamics of the load and manipulator itself must be taken into account. 1 In a mobile manipulator with lighter weight and higher motion speed, the existence of the flexible structure on joints and links degrades the motion accuracy. Accordingly, an additional constraint must be imposed, when the DLCC is determined for flexible mobile manipulators. The constraint should limit the end effector tracking error due to the deflection or oscillation of the mobile manipulator. Wang and Ravani 1,2 presented an algorithm to determine the load carrying capacity on fixed base robots. The limitation of the rigid joint and link assumption in the formulation and analysis of flexible manipulators were investigated by the other researchers. 37 This paper, presents a new method to determine the load carrying capacity for mobile manipulators with flexible links and joints during the motion between two end positions. For an analysis of flexibility, the Lagrangian assumed mode is employed. Then, the problem of finding the maximum load carrying capacity for flexible mobile manipulators is converted into a trajectory optimization problem. An objective function is defined and dynamic equations are * Corresponding author. Inteligent Systems Laboratory, Faculty of Mechanical Engineering, Zanjan University, Zanjan (Iran). E-mail: ghariblu@mail.znu.ac.ir Mechanical Engineering Department, Iran University of Science and Technology, Tehran (Iran). linearized, as well as other constraints on its state space form to solve with Iterative linear Programming (ILP) method. II. LINEARIZED STATE SPACE REPRESENTATION OF DYNAMIC EQUATIONS The Lagrangian assumed mode formulation is used to model the dynamics of the link flexibility on a mobile manipulator 8 (Figure 1). Also, the elastic mechanical coupling between the i th joint and link is modeled as a linear torsional spring with a constant stiffness coefficient k ti (Figure 2). Then, the compact form of extended equation of motion, 9 can be represented as: [M] e ¨ q e + C( q e , ˙ q e ) + G( q e ) + [K t ] e [q ] e = τ e (1) If the load is treated as a variable, the dynamic Equation (1) is rearranged as: ¨ q e = [M( q e ,m L )] 1 e ( τ e C( q e , ˙ q e ,m L ) G( q e ,m L ) [K t ] e q e ) = f ( q e , ˙ q e , τ e ,m L ) (2) Defining the state vector X = [x 1 x 2 ] T , where x 1 = q e and x 2 = ˙ q e , the state space representation of Equation (2) can be rewritten in the following form: ˙ X = ˙ x 1 ˙ x 2 = x 2 f ( X(j ), τ e (j ),m L ) = F ( X(j ), τ e (j ),m L ) (3) Discretized form of the equation (3) is: X(j + 1) X(j ) h = F ( X(j ), τ e (j ),m L ), j = 1, 2,...,p (4) where, h = T p and T is the overall motion time and p is the total number of time intervals. The nonlinear function f at the (k + 1)th trajectory is expanded in Taylor series about the kth trajectory. After neglecting the higher order (non linear) terms the following equation is obtained in a matrix form: X(j + 1) = [G j ] X(j ) + [H j ] τ e (j ) + B j m L + D j (5) The elements of matrices [G j ], [H j ] and B j , D j are found in Reference [10]. X(j + 1) can be written as a linear