Programming Language Support for Multisensor Data Fusion: The Splash Approach

Abstract

We present the Splash programming framework to support the effective implementation of multisensor data fusion. Multisensor data fusion has been widely exploited in autonomous machines since it outperforms algorithms using only a single sensor, in terms of accuracy, reliability and robustness. Knowing that developers have long been lacking programming language support for multisensor data fusion, we offer a dedicated Splash language construct along with formal semantics for multisensor data fusion. Specifically, we analyze the structural characteristics of multisensor data fusion algorithms and derive technical issues that the language construct must tackle. We then give a detailed account of the language construct along with its formal semantics. Finally, we validate its utility and effectiveness via its application to a lane keeping assist system.