@InProceedings{Hazra95, title = "{A}pplication of {T}ransputer-based {P}arallel {C}omputation in {M}atching {R}eal-{T}ime {C}ontrol {M}odels", author= "Hazra, Tushar K.", editor= "Nixon, Patrick", pages = "197--212", booktitle= "{P}roceedings of {W}o{TUG}-18: {T}ransputer and occam {D}evelopments", isbn= "90 5199 222 X", year= "1995", month= "mar", abstract= "Mathematical models describing physical processes can be 'matched' to discrete samples of processes by considering measured input-output data and adjusting coefficient parameters of the model to provide an optimal agreement between the system and model responses. This well known technique has been widely implemented in an off-line basis for many years. The process of matching can be based on standard optimization procedures such as Davidon Fletcher Powell, Fletcher Reeves or even simply Newton Raphson methods. Standard simulation routines such as Runge Kutta or Euler's methods can be used to generate the model responses. However, the procedure for model matching involves intensive computing. This paper offers a number of prospective approaches to the solution based on parallel computing methodologies. Using standardized model matching experiments and simulated noise-free data, performance comparisons for sequential and multiprocessor computations have been evaluated on a transputer-based system." }