Aiming at the two problems of poor real-time performance and calculation accuracy caused by insufficient processor computing power and limited effective bits when implementing traditional integrated navigation simulation algorithms on hardware platforms built with MEMS inertial devices and industrial-grade processors, this study quantifies the noise and error magnitudes of MEMS inertial devices, simplifies Kalman filter error equations, and adopts UD decomposition algorithm. The optimized algorithm achieves 17% improvement in computational speed, effectively demonstrating its validity. Ultimately it is implemented on a low-cost attitude measurement system composed of MEMS devices. Vehicle tests verify that the heading angle accuracy reaches 0.5° and horizontal attitude angle accuracy reaches 0.2°, successfully resolving the issues of real-time navigation data output and computational precision.