Automatic Denavit-Hartenberg Parameter Identification for Serial Manipulators Carlos Faria DTx-Colab, Centre Algorithmi and Department of Industrial Electronics University of Minho Guimar˜ aes, Portugal cfaria@dei.uminho.pt Jo˜ ao L. Vilac ¸a 2Ai Polytechnic Institute of C´ avado and Ave Barcelos, Portugal jvilaca@ipca.pt S´ ergio Monteiro Centre Algorithmi and Department of Industrial Electronics University of Minho Guimar˜ aes, Portugal sergio@dei.uminho.pt Wolfram Erlhagen Centre of Mathematics and Department of Mathematics University of Minho Guimar˜ aes, Portugal wolfram.erlhagen@math.uminho.pt Estela Bicho Centre Algorithmi and Department of Industrial Electronics University of Minho Guimar˜ aes, Portugal estela.bicho@dei.uminho.pt Abstract—An automatic algorithm to identify Standard Denavit-Hartenberg parameters of serial manipulators is pro- posed. The method is based on geometric operations and dual vector algebra to process and determine the relative transforma- tion matrices, from which it is computed the Standard Denavit- Hartenberg (DH) parameters (ai , αi , di , θi ). The algorithm was tested in several serial robotic manipulators with varying kinematic structures and joint types: the KUKA LBR iiwa R800, the Rethink Robotics Sawyer, the ABB IRB 140, the Universal Robots UR3, the KINOVA MICO, and the Omron Cobra 650. For all these robotic manipulators, the proposed algorithm was capable of correctly identifying a set of DH parameters. The algorithm source code as well as the test scenarios are publicly available. Index Terms—Kinematic identification, Denavit-Hartenberg parameters I. I NTRODUCTION Kinematic identification refers to the determination of a minimal number of parameters that completely describe the position and orientation of a manipulator’s structure as a func- tion of its joint positions. Many models have been proposed to characterize a kinematic structure: the Standard and Modified Denavit-Hartenberg (DH) convention [1], the Hayati model [2], the Stone and Sanderson’s S-model [3] and recent models based on Product of Exponentials (POE) [4], [5]. The Denavit-Hartenberg model is still the most used con- vention to represent the robot’s kinematic structure. It provides a guaranteed minimal representation, an intuitive method to determine its parameters and most importantly, it works on straight-forward linear algebra whose matrices are computa- tionally fast to solve. This work has been supported by FCT – Fundac ¸˜ ao para a Ciˆ encia e Tec- nologia within the Project Scope: UID/CEC/00319/2019, the FCT scholarship grant: SFRH/BD/86499/2012 and the DTx-Colab. Independent of the convention used, there is a consistent problem with robots, particularly serial structures that causes repeatable but inaccurate movements. This problem derives from manufacturing and assembly tolerances, wear and tear, or permanent bending due to fatigue. The listed error sources are reflected on the real kinematic model parameters, and on the gap to the nominal parameter values. Kinematic errors are especially impactful in serial manipulators where the parameter deviations propagate through the kinematic chain. Industrial robot calibration methods, particularly the ones based on kinematic parameters, are compartmentalized in four steps, i) modelling, ii) measurement, iii) identification, and compensation [6], [7]. The proposed algorithm is primarily related to the parameter identification step as a sub-type of planar calibration methods [8], [9]. Another common problem to users that need to model kinematic structures of robotic manipulators relates to the inaccessibility of the model parameters, which are usually han- dled internally by the controller. Even if the robot is properly calibrated, the user has no access to the kinematic parameters other than the nominal values in the documentation. In this paper, we propose an algorithm for DH parameter identification based on geometry and dual vector algebra for any type of serial robot. The algorithm splits into two parts, the first called “feature identification”. In this part, the robot’s end-effector position is acquired after sequential movements in each joint. The acquired set of points are processed to determine the motion axis of each joint, an idea originally explored by Stone [3] to determine the S-model parameters. The second part, “parameter extraction”, applies dual vector algebra to calculate the intermediate coordinate frames be- tween consecutive joints, and then to extrapolate the Standard DH parameters. The dual vector algebra concept was applied 978-1-7281-4878-6/19/$31.00 ©2019 IEEE 610 Authorized licensed use limited to: b-on: UNIVERSIDADE DO MINHO. Downloaded on January 18,2021 at 23:31:11 UTC from IEEE Xplore. Restrictions apply.