G. Robert Redinbo

Professor Redinbo’s research is in the general area of the design and analysis of reliable computing resources, both at the digital system and subsystem levels. His work emphasizes fault-tolerant computer implementations and is motivated by theoretical concepts as well as practical experience. He is particularly interested in communication and signal processing functions that are affected by soft errors causing momentary malfunctions. Earth-orbiting satellites containing high-speed processing resources which support fast transforms and multiplexing operations are studied. Data compression procedures which are very susceptible to processing failures benefit from reliable design features. JPEG and arithmetic coding data compression standards are of particular focus. A distinctly new approach to reliable computing uses algorithmic-level fault tolerance techniques for increasing dependability in processed data by employing real number error-detecting codes. An important challenge is separating numerical roundoff effects from internal hardware failure effects. Professor Redinbo also studies real number error-detecting and correcting codes, their design and error correction procedures. Such codes involve symbols that have real or integer values as opposed to classic binary codes. The new, developing wavelet codes which are similar to convolutional codes hold great promise of protecting many data processing subsystems.