Pervasive simulation envisions to apply computationally expensive simulations in everyday scenarios on resource-constrained mobile devices. The PerSiVal project aims to realize a bio-mechanical simulation of the human arm as augmented reality application. To enable the execution of such computationally expensive simulations on resource-constrained devices surrogate models are applied. Nonetheless, the execution of the surrogate models can still be challenging. A solution to deal with the constraint resources is the offloading of the surrogate model to a server that is wirelessly connected to the mobile device. Challenges arise due to inevitable delays caused by processing and communication between devices. To counterweight the delays, previous work has applied a second, light-weight surrogate model with lower performance on the mobile device. The goal of this thesis is the design and evaluation of Kalman filter-based approaches for fusion in the presence of delays, jitter and losses. This work contributes an improved strategy for surrogate model derivation, improved and light-weight surrogate models for the muscle simulation, a distribution model for reproducible analysis and evaluation of distributed algorithms, an improved variant of the fusion algorithm from previous work and the design and evaluation of Kalman filter-based solutions to the fusion problem. While being computationally more demanding, the Kalman filter-based approaches show a significant advantage in dealing with delays, jitter and losses in the evaluated scenarios. Especially the constant velocity Kalman filter with input augmentation fusion works best in the tested scenarios. Conclusively, the Kalman filter is a powerful framework that has been successfully applied in the context of distributed pervasive simulations for continuous problems.