Modeling and Control of Elastic Joint Robots

From MaRDI portal





It has been recognized that joint elasticity is the main source of vibrational behaviour in current industrial robot arms whose links are rigid. For such a class of robot manipulators, this paper presents a simplified dynamic modeling not considering the gyroscopic couplings between the actuating motors and the driven links; this results in neglecting some off-diagonal terms in the system inertia matrix. The assumption is reasonable when the gear ratios are large. From the control point of view, robots with joint elasticity display basic different properties w.r.t. fully rigid manipulators. In general, the complete form of the dynamic model cannot be exactly linearized using nonlinear static state-feedback. Moreover, when the system outputs are the link positions, the input-output decoupling problem is not solvable using this type of feedback laws. In these cases, nonlinear feedback of the dynamic type is required. However, based on the proposed simplified model, it is shown here that any robot arm with elastic joints can be feedback linearized and decoupled using only static state-feedback. The resulting closed-loop system is described by strings of four integrators, as opposed to the structure of two obtained in the rigid case; the synthesis of a tracking controller is then easily completed. The robustness of this control law is also investigated. For weakly elastic joints, an altenative approach is detailed based on the integral manifold concept. The dynamic equations can be recasted as those of a singularly perturbed model. A feedback control law is then found which renders invariant a proper submanifold in the state-space. The advantage of this scheme is that, when damping is added at the joints, state measurements are needed only at one side of the joint elasticity. The integral manifold approach provides a completely computable control law only when applied to a system which is feedback linearizable by static feedback. It is interesting to compare this result with those of \textit{M. W. Spong} and \textit{M. Vidyasagar} [IEEE J. Robot. Autom. RA-3, 345-351 (1987)]; when working with the full model, only an approximation of the invariant feedback law can be computed. The paper is well-written and mathematically sound. It has the merit of collecting various mainstream ideas in this active field. Only two points deserve some criticisms. First, the example used in section 2 to motivate the approximated computation of inertia in the model is not pertinent (e.g. for planar motion). Second, the reported simulations refer only to a single-link elastic-joint arm: this is a case where the model is feedback linearizable via static feedback even without the simplifying assumption.




Cited in
(only showing first 100 items - show all)








This page was built for publication: Modeling and Control of Elastic Joint Robots

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3804564)