A Nonlinear Observer Approach to Diagonally Decoupled Direct Visual Servo Control

Abstract

This brief investigates the problem of vision-based robot control where the equilibrium is defined via a reference image. Specifically, this work considers the class of intensity-based nonmetric solutions, which provide for high accuracy, versatility, and robustness. The existing general techniques within that class present either a fully coupled control error dynamics or at best only achieve decoupling of the translational part, i.e., they can only obtain a triangular system in the general case. These couplings in the system dynamics increase analysis complexity and may degrade system performance. This work proposes a new nonlinear observer-based strategy for completely decoupling the translational and rotational parts, i.e., to obtain a diagonal system in the general case. A Lyapunov-based analysis of local stability and convergence, as well as proofs of diffeomorphism and of that decoupling property are provided. Improved performances are also experimentally confirmed using a camera-mounted six-degree-of-freedom (DoF) robotic manipulator in a challenging setup. In particular, execution times are drastically reduced by using the proposed diagonal technique.