Fault tolerance is increasingly important for space and hazardous environment robotics. Robots need to quickly detect and tolerate internal failures in order to continue performing their tasks without the need for immediate human intervention. Using analytical redundancy, this paper derives tests with which the robot can detect failures. The paper also develops a layered intelligent control framework containing these new sensor-based fault detection and tolerance schemes. the servo, interface, and supervisor layers form a hierarchy of fault tolerance which provide different levels of detection and tolerance capabilities for structurally diverse robots.