Optimising multirate motion control algorithms on parallel processors by static scheduling
Ronald G. Harley, M.J. van der Westhuizen, David Levy, D.R. Woodward · 2002
The use of computational intensive adaptive control algorithms is often limited for systems requiring high sampling rates. Most modern microprocessors are stretched to the limit to perform all the required calculations within the short sample times required for such applications as the control of AC machines with neural networks. Microprocessors such as the Texas Instruments C40 and the Inmos transputer family provide the option of using parallelism so as to share the computational load amongst a number of processors. However, efficient programming of multirate algorithms on parallel processor networks is a complicated task which entails combinatorial optimisation. This paper reports on the development of a general scheduling tool which is used to implement complex control algorithms on parallel processors by merely drawing the required block diagram with standard schematic software. The scheduling tool then provides accurate estimates of the sampling times achievable for a given algorithm and also generates the required C language code. Results of this scheduling tool are presented showing time-space graphs (Gantt charts) for typical control algorithms.>