VIF will evaluate the comparative strengths and weaknesses of different drone sensor modalities for sensor fusion. Different sensor sets and their specifications shall be defined for the use cases based on the mission requirements. KPIs for sensor data fusion algorithms shall be defined.
VIF will set up a simulation architecture including dynamic drone models, sensor models (realistic sensor behaviour depending on weather, light conditions etc.) and environment modelling. VIF will develop AI-based/conventional algorithms for sensor data fusion. This shall include ego-motion estimation and localization, object detection and classification, free space estimation and environmental mapping (e.g., SfM, multi-view stereo, Bayesian mapping). VIF will demonstrate the developed sensor fusion algorithms and use cases in established virtual simulation architecture and demonstrate the sensor fusion algorithms on recorded drone data.