A Cross-Platform Network-Ready Visual Servo Simulator

The field of visual servoing is now widely accepted as a modern, consolidated discipline for vision-based real-time robot control. Though experimental setups have become affordable, theoretical and practical issues arise, which advise the use of simulation systems prior to real execution. This paper presents an object-oriented, cross-platform, network-ready environment for visual servoing simulations. With flexibility and extensibility as the main design goals, tasks can be defined either for cameras attached to moving Cartesian frames, or serial link manipulators. In order to provide fast feedback to the user, it includes real-time 3D rendering of the simulated scene. Task parameters and visual features can be freely chosen, and new features can be easily added to the framework. Output data is logged to disk files, which can be analyzed by any popular mathematical package. The simulator is built upon an agent-based framework, which makes possible the distribution across multiple networked computers. Moreover, it can be securely downloaded from a Web server and automatically installed in a computer, running either Windows, Linux or Mac OS X operating systems. This simulator is now a valuable tool in our research lab, and it has been extensively used by the students of a recent online course on visual servoing