The next generation of robotic applications--computer integrated manufacturing, teleoperation, and mobile autonomous robots--will require far more computer systems support than currently available. In particular, real-time supervisory control systems will be needed to integrate an increasing number of sensors and actuators, as well as to communicate with other computers in a distributed environment. This thesis describes the design and implementation of SAGE, an operating system built specifically for real-time robotic supervisory control. The SAGE kernel runs on off-the-shelf Motorola 68020 processor boards, and features lightweight processes, virtual memory support, extensible low-overhead synchronization primitives, and real-time communications capabilities. Because SAGE is one of the first systems built for robotic supervisory control, the thesis focuses on the issues and design tradeoffs that arise in building a supervisory control operating system. The thesis also describes how SAGE was used to control a number of intelligent devices, including a Utah/MIT hand and a PUMA robot arm. The robotic experiments performed demonstrate that the operating system can be used in real-time supervisory control applications.