On the generation of realistic simulated inertial measurements

Abstract

The aim of the research presented in this paper was to develop a software component simulating the output of an Inertial Measurement Unit (IMU) in a wide range of environments. The requirements to fulfil by such a software tool were carefully defined, which led to a system architecture that put especial emphasis on delivering, at the same time, a tool general enough as to cope gracefully with innovation - that is, sensor evolution - as well as being able to be extended at almost no cost. A rigorous data modeling was the key tool to achieve the aforementioned goals; the main data entities involved in the generation of error data (noise) were identified and characterized. Different sources of noise may then be implemented using an Object Oriented approach (general software) and (dynamically loadable) shared libraries (extensibility). The mathematical background used to simulate the different sources of error (noise) data is also presented, as well the different strategies employed to validate the results of the IMU simulator. The tool has been successfully validated for pedestrian, terrestrial and airborne scenarios through stochastic analysis tools.