The Agile Eye is a 3-DOF 3-RRR spherical parallel manipulator developed for the rapid orientation of a camera. Its mechanical architecture leads to high velocities and accelerations. First, the kinematic model of this manipulator was developed. Then, a geometric optimization was carried out in order to determine the dimensional parameters which would produce the best accuracy for the mechanism. A complete dynamic model was then established. Finally, a prototype was designed and built, and a high-performance controller based on a DSP was developed. The prototype was built in 1993 and has been gaining in popularity ever since.

The workspace of the Agile Eye is superior to that of the human eye. The miniature camera attached to the end-effector can be pointed in a cone of vision of 140° with ±30° in torsion. Moreover, due to its low inertia and its inherent stiffness, the mechanism can achieve angular velocities above 1000°/sec and angular accelerations greater than 20000°/sec² which is beyond the capabilities of the human eye.

One of the most interesting topics of research related to the Agile Eye is the analysis of its singularities. Surprisingly, the singularity loci of the Agile Eye are independent from the chosen branch (there are a total of 8 branches). Note that for general 3-RRR spherical parallel manipulators, the singularity loci are strictly dependent on the chosen branch. In addition, in the Agile Eye, there exist four poses for the mobile platform in which arbitrary finite motions of the actuators do not produce any output at the mobile platform. Finally, the direct kinematic problem of the Agile Eye allows 8 assembly modes.

A simplified 2-DOF version of the Agile Eye has also been recently developed and patented. Its obvious advantages are very low cost, trivial inverse kinematics, simplified direct kinematics (only four assembly modes), enlarged workspace, and trivial singularity analysis. Another interesting feature is that the orientation of the camera (the mobile platform) is completely specified by an azimuth and an altitude angle, i.e., there is no torsion.

This architecture was also used, in collaboration with the Computer Vision and Digital Systems Laboratory (CVDSL), for the development of an agile stereo pair, involving two rapid, precise and independent cameras enabling three-dimensional vision.