This final report presents research activities and results within the project Autonomous surveillance with collaborating sensors (2018-2020). The project has investigated and evaluated how multi-sensor systems of collaborative mobile and stationary units should be designed to autonomously detect, classify and monitor threats in militarily challenging surveillance situations. The multi-sensor architecture for a system that will achieve robustness and endurance when monitoring military installations has been designed at a high level. The central functions have been identified and functional connections have been partly described. For the flying multi-sensor platforms, two different positioning methods have been investigated that enable autonomy without satellite navigation information. Real-time positioning based on image data has been developed, and positioning based on magnetic field measurements has been evaluated. Methods for detection and tracking of moving objects have also been developed. A framework for evaluating new methods for distributed target tracking has been developed. To study the performance of moving platforms with dynamic navigation uncertainty, target tracking results from a simulation with a UAV have been evaluated. The results show that the navigation uncertainty must be taken into account in target tracking. A new threat detector based on deep learning has been developed. The purpose is to detect various threats in the military surveillance arena. The threat detector has been validated on the basis of data from a simulation model. The results show that the method has the ability to detect the various threats.