Community Research and Development Information Service - CORDIS

Optimal control of robots: from fiction to reality

Robotic systems (RS) constitute specialised devices designed to perform a variety of functions that are normally ascribed to humans. The dynamic performance of RS distinctly relates to its effective control for which a number of properly identified parameters are necessary. A Bulgarian academic institution succeeded in developing efficient methods for the accurate estimation of these dynamic parameters for designing robotic systems with robust control.
Optimal control of robots: from fiction to reality
The first modern robot was developed in the late 50's and was purposed mainly for simple industrial applications, such as for transferring parts. Since then, RS have been further improved and today they can perform more complex tasks, including surgeries. For the successful completion of such delicate duties, their dynamic performance needs to be both carefully and accurately controlled.

In order to achieve optimal control of RS, it is necessary to accurately describe a set of parameters associated with its highly complicated dynamics. This is often quite difficult to accomplish and involves considerable time and cost for the proper identification of these dynamic parameters. To resolve this, a new comprehensive approach has been used for the effective modelling of control mechanisms of RS.

With the new approach, the required dynamic performance of RS reflects a set of controlled outputs highly affected by driving inputs. Including key terms, such as inertia, driving, velocity and gravity, a system of mathematical equations represents the full lumped parameter model of input-output relationships. The new approach is suitable for both accurate parameter identification and robust control design.

The models have been verified on various RS, such as a robot manipulator, a mobile (wheeled, legged, underwater) robot, and a biped robot. These RS were equipped with various-type actuators including electrical motors, electro-hydraulic cylinders, and artificial muscles. The new methods feature simplicity and ease of use, while they lead to a significant reduction of errors, improvement of accuracy and minimisation of the required time/cost.

Furthermore, this approach is useful for selecting the most appropriate potential trade-off solutions between efforts and costs for the interlinked problems of full dynamic modelling. Not only does it contribute to the accurate parameter identification and optimal control design, but also to the optimisation of any problem with RS performance. The institute is looking for partners to support these methods for further development.

Related information

Record Number: 80725 / Last updated on: 2005-09-18
Domain: IT, Telecommunications