Physical phenomena, such as high-frequency vibration, time delay, actuator and sensor nonlinear characteristics, contamination of signals by noise, are usually eliminated by model simplification when control schemes are synthesized. Therefore, experimental studies play an important role in control system design in the verification of new control syntheses in order to improve them for practical applications, in particular, when the above phenomena can not be neglected. Such experimental studies are based on highly complex electro-mechanical systems with muld-sensors and multi-actuators and advanced software environment. Therefore, the control schemes tend to be sophisticated, and such experimental systems are both I/O and computationally intensive.

This thesis focuses on system design and implementation of real-time control hardware and software architectures for three experimental facilities which were developed as test beds for verification and experimental studies of new control schemes. The three experimental facilities are:​ (i) a master-slave telerobot system (MST) for bilateral control;​ (ii) a coordinated dual-arm system (CDAS) for adaptive force and motion control;​ and (iii) a multi-arm reconfigurable facility (RoboTwins) for grasping and manipulation studies. These experimental setups use complex real-time control systems. They are characterized by modular and open hardware and software architectures that can be easily reconfigured, and each control level be directly accessed.

In this work, a systematic study of computer hardware suitable for real-time control applications is provided. The goal is to achieve well-designed system architectures, full utilization of hardware resources, and specified performance of control systems. It is clear that efficient hardware architectures provide a prerequisite for building reliable and user-friendly real-time software architectures and for consistent integration of hardware and software.

The primary criteria for the hardware architecture for an experimental facility are modularity, expandability, and system reconfigurability. In addition, an adequate computational resource is needed for high bandwidth and high performance real-time control. Consequently, hardware structure for such experimental robotic systems are mostly multiprocessor-based. Two architectures were designed and implemented. One is a VMEbus-based architecture which is implemented for the telerobot system;​ the other is a two-level distributed-parallel multiprocessor architecture which was implemented for the RoboTwin and the CDAS facilities.

In this work, a versatile real-time software environment (RTSE) for the three setups has been designed, based on a multiprocessor resource-limited hardware configuration. The real-time software is characterized by a modular multilayer, user-friendly, and easily maintained architecture. The software consists of the following hierarchical levels:​ an extended-computer module level (low-level control), a real-time operating system (RTOS), an abstract robot control level, and a user integration level. At the user integration level, the software is not only independent of hardware and the RTOS, but also easily reconfigurable, based on basic software functional modules which were built at the abstract robot control level. The real-time software environment has been used on the three experimental facilities for verification and experimental studies of advanced control schemes.