Constrained visual servoing under uncertain dynamics

Abstract

In visual servoing, limitations in the field of view of the vision sensors are either ignored or treated at the kinematic level. The former can easily jeopardise task success, while the latter reduces the maximum achievable robotic motion speeds. In this work, the aforementioned literature gap is filled by designing and rigorously analysing a torque controller that guarantees prescribed transient and steady-state performance attributes on the image feature coordinate errors, while respecting the field-of-view constraints. No path planning and no information regarding the actual system dynamics are required. In addition, no approximation structures (i.e. neural networks, fuzzy systems, etc.) are utilised to acquire such knowledge. The proposed visual servo controller is static, involving very few and simple calculations to produce the control signal, making its implementation on embedded control platforms straightforward. Simulation studies are utilised to illustrate the motivation and to clarify–verify the theoretical findings.