Remote maintenance of real time controller software over the internet

The aim of the work reported in this thesis is to investigate how to establish a standard platform for remote maintenance of controller software, which provides remote monitoring, remote fault identification and remote performance recovery services for geographically distributed controller software over the Internet. A Linear Quadratic Gaussian (LQG) controller is used as the benchmark for the control performance assessment; the LQG benchmark variances are estimated based on the Lyapunov equation and subspace matrices. The LQG controller is also utilized as the reference model of the actual controller to detect the controller failures. Discrepancies between control signals of the LQG and the actual controller are employed to a General Likelihood Ratio (GLR) test and the controller failure detection is characterized to detect sudden jumping points in the mean or variance of the discrepancies. To restore the degraded control performance caused by the controller failures, a compensator is designed and inserted into the post-fault control loop, which serially links with the faulty controller and recovers the degraded control performance into an acceptable range. Techniques of controller performance monitoring, controller failure detection and maintenance are extended into the Internet environment. An Internet-based maintenance system for controller software is developed, which provides remote control performance assessment and recovery services, and remote fault identification service over the Internet for the geographically distributed controller software. The integration between the mobile agent technology and the controller software maintenance is investigated. A mobile agent based controller software maintenance system is established; the mobile agent structure is designed to be flexible and the travelling agents can be remotely updated over the Internet. Also, the issue of heavy