The eigenstructure assignment scheme for robust multivariable feedback control is extended to redundantly actuated dynamical systems. It is shown that an orthonormal set of close loop eigenvectors is always exactly assignable in the case of redundant actuation proving the inherent robustness in the control design methodology. A choice of close loop eigenvector set to minimize the feedback gain matrix is suggested. Partial Eigenstructure Assignment methodology is proposed for second order mechanical systems. A methodology for coordinated actuation of redundant actuator sets by a trained weighted minimum norm solution is presented. To apply the methodology to hyper-redundant actuator arrays, for application to smart actuator arrays, a novel adaptive discretization algorithm is proposed. The adaptive aggregation strategy, based on the physics of the system, introduces nodes, to optimize a performance index of the overall plant model. The dimensionality of the inputs thus reduces to a finite number, making it a candidate plant for control by the robust redundant control scheme. The adaptive aggregation together with robust redundant control methodology is demonstrated on a finite element model of a novel morphing wing. This schema unifies the traditionally disparate methods of modeling and controller design.

Redundantly actuated manipulators have only recently aroused significant scientific interest. Their advantages in terms of enlarged workspace, higher payload ratio and better manipulability with respect to non-redundantly actuated systems explain the appearance of numerous applications in various fields: high-precision machining, fault-tolerant manipulators, transport and outer-space applications, surgical operation assistance, etc. The present Ph.D. research proposes a unified approach for modeling and actuation of redundantly actuated parallel manipulators. The approach takes advantage of the actuator redundancy principles and thus allows for following trajectories that contain parallel (force) singularities, and for eliminating the negative effect of the latter. As a first step of the approach, parallel manipulator kinematic and dynamic models are generated and treated in such a way that they do not suffer from kinematic loop closure numeric problems. Using symbolic models based on the multibody formalism and a Newton-Euler recursive computation scheme, faster-than-real-time computer simulations can thus be achieved. Further, an original piecewise actuation strategy is applied to the manipulators in order to eliminate singularity effects during their motion. Depending on the manipulator and the trajectories to be followed, this strategy results in non-redundant or redundant actuation solutions that satisfy actuator performance limits and additional optimality criteria. Finally, a validation of the theoretical results and the redundant actuation benefits is performed on the basis of well-known control algorithms applied on two parallel manipulators of different complexity. This is done both by means of computer simulations and experimental runs on a prototype designed at the Center for Research in Mechatronics of the UCL. The advantages of the actuator redundancy of parallel manipulators with respect to the elimination of singularity effects during motion and the actuator load optimization are thus confirmed (virtually and experimentally) and highlighted thanks to the proposed approach for modeling, simulation and control.