PARALLEL HIGH-PRECISION ORBIT PROPAGATION USING THE MODIFIED PICARD-CHEBYSHEV METHOD

The modified Picard-Chebyshev method, when run in parallel, is thought to be more accurate and faster than the most efficient sequential numerical integration techniques when applied to orbit propagation problems. Previous experiments have shown that the modified Pica rd-Chebyshev method can have up to an order of magnitude speedup over the 12th order Runge-Kutta-Nystrom method. For this study, the evaluation of the accuracy and computational time of the modified Picard -Chebyshev method, using the Java Astrodynamics Toolkit (JAT) high-precision force model, is conducted to assess its runtime performance. Simulation results of the modified PicardChebyshev method, implemented in MATLAB and the MATLAB Parallel Computing Toolbox, are compared against the most efficient first and second order Ordinary Differential Equation (ODE) solvers. A total of six processors were used to assess the runtime performance of the modified Picard-Chebyshev method. It was found that for all orbit propagation test cases, where the gravity model was simulated to be of higher degree and order (10 additional function calls to JAT using a 70 degree x 70 order Earth Gravity Model to increase computational overhead to 0.142 seconds per force function call), the modified PicardChebyshev method was faster, by as much as 100%, than the other ODE solvers which were tested.