UAV Environment Sensing And Collision Avoidance

With the gradual popularization and application of UAV in the military and civilian fields,the future airspace will become increasingly crowded,presenting a complex air traffic situation shared by both manned and unmanned aircrafts,which will undoubtedly bring huge safety issues to the existing air traffic system.The Sense and Avoid((SAA)is the core technology ensuring UAV safety operation in civil airspace and prerequisite of UAV’s autonomy and intelligence capability.The SAA function is to effectively observe,evaluate the operating environment through sensory and communication data,then to avoid collisions by generating effective evasive paths and maneuver control.This thesis mainly focuses on the three core problems,namely the sensing,planning and control of SAA,as follows:1.A SAA systematic framework featured by the safety,accuracy and efficiency is porposed.First,a unified joint optimization is constructed with regard to the three main ingredients of SAA,which are the sensing,planning and control.Taken into consideration of functional and missionary difference between different UAV platforms,two separated set of technical standards and system reference architectures are proposed to guide the following algorithmic and systematic research.2.A monocular vision based SAA method is proposed considering its payload limitation.First,a 2D safety envelop is defined based on the relative angle to model the threat level of collisions,then a visual servoing based reactive controller is designed to guide the collision avoidance maneuver.The effectiveness of the proposed monocular vision based SAA method is validated with both simulations and experiment.3.A collaborative sensing method is proposed for the target estimation in multiple UAV application scenario.Specifically,a distributed hybrid information filter is designed with two different linearization techniques.The sufficient conditions for the system stability are also investigated.The proposed method requires only one communication iteration between every two consecutive time instants.Both simulations and experiments are extensively studied to show the performance of the proposed method.4.A multi-constrained distributed controller is designed to guide the coordination of UAV flocking.Specially,a distributed model predictive controller(DMPC)combined with the consensus base ADMM is proposed incorporating collision avoidance,state and control input constraints.Further,the sufficient conditions for system feasibility and stability are provided under limited ADMM iterations.The proposed method is able to realize comparative performance to the centralized method.Simulation and experiment results show that the DMPC-ADMM method can fulfill the flocking control task under practical constraints.5.A set of SAA system integration and SAA experimental verification is described.For the high cost large and medium-sized UAVs,a software based simulation system is developed to test various sensors,algorithms and platform characteristic on different SAA scenarios.For the small and micro UAVs,various systems have been developed to test the uncooperative/cooperative SAA with regard to dynamic/static obstacles.