Dynamic Identification of the Franka Emika Panda Robot With Retrieval of Feasible Parameters Using Penalty-Based Optimization

• Published in: IEEE robotics and automation letters Vol. 4; no. 4; pp. 4147 - 4154
• Main Authors: Gaz, Claudio, Cognetti, Marco, Oliva, Alexander, Robuffo Giordano, Paolo, De Luca, Alessandro
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.10.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Computer Science; Computer simulation; Control theory; dynamic identification; Dynamic models; Dynamics; Economic models; feasible physical parameters; Franka Emika Panda; friction model; Inertia; Inverse dynamics; Manipulator dynamics; nonlinear global optimization; Optimization; Parameter identification; penalty methods; Robot arms; Robot control; Robot sensing systems; Robotics; Robots; Service robots; Tensors; Torque
• ISSN: 2377-3766; 2377-3766
• DOI: 10.1109/LRA.2019.2931248

In this letter, we address the problem of extracting a feasible set of dynamic parameters characterizing the dynamics of a robot manipulator. We start by identifying through an ordinary least squares approach the dynamic coefficients that linearly parametrize the model. From these, we retrieve a set of feasible link parameters (mass, position of center of mass, inertia) that is fundamental for more realistic dynamic simulations or when implementing in real time robot control laws using recursive NewtonEuler algorithms. The resulting problem is solved by means of an optimization method that incorporates constraints on the physical consistency of the dynamic parameters, including the triangle inequality of the link inertia tensors as well as other user-defined, possibly nonlinear constraints. The approach is developed for the increasingly popular Panda robot by Franka Emika, identifying for the first time its dynamic coefficients, an accurate joint friction model, and a set of feasible dynamic parameters. Validation of the identified dynamic model and of the retrieved feasible parameters is presented for the inverse dynamics problem using, respectively, a Lagrangian approach and Newton-Euler computations.