Cooperative Collision Avoidance Method For Multi-UAV Systems

Due to the load and the endurance of a UAV is limited generally,and the multiple UAVs construct a cooperative system have benefits in terms of practicability and robustness,multi-UAV cooperation is an important direction of UAV system development and application in the future.However,with the widespread use of consumer UAVs,the development of multi-UAV formation flights,and the emergence of multi-UAV cooperations,the flight safety situation of UAVs is becoming more and more severe,and the possibility of collision between UAVs in the same airspace has increased dramatically.Collision avoidance is one of the basic requirement of the UAV system,and it is an important guarantee for the UAV to fly safely and perform other high level tasksThe existing collision avoidance methods are mainly categorized into two categories Model-based methods and Model-free methods.Although Model-based methods often re-quire high computational costs,security is usually guaranteed.Model-free enables real-time response,but does not provide security guarantees.Multi-UAV collision avoidance problem researched in this thesis needs to ensure safety and need to match real-time performance For above reasons,we propose an real-time hybrid fully distributed cooperative collision avoidance method based on the velocity obstacle method of model-based class and the fuzzy control method of model-free classIn order to ensure that the structure of the rules in the fuzzy control does not change with the number of UAVs in a system,we partition the sensing region of the UAV into different collision regions firstly.Within each collision region,from the "intruders" who entered the sensing range,the" most threatening" who needs to be addressed urgently is selected through collision check and collision risk assessment.After that,the fuzzy control with good real-time performance calculates a candidate velocity based on the status information of these"most threatening".However,this candidate velocity does not guarantee that collision-free However,the candidate velocity can not guarantee collision-free,so this thesis checks and optimizes it,and sends it to the UAV for execution.Finally,we build a simulation platform for validating the proposed method,and compare it with the pure fuzzy control method in different dimensions such as the length of the traversed trajectory and the used time to illustrate the effectiveness of the proposed method The experimental results illustrate that our approach can improve the collision avoidance performance of the UAV immensely,and can effectively solve the potential collisions between UAVs and ensure flight safety.In our propose method,at each moment,since the own UAV only needs to communicate with its surrounding UAVs to obtain their state information,the method is fully distributed.Each drone only needs to broadcast its own state information in real time,and obtain state information of other UAVs when needed,so the communication of the system is asynchronousThe thesis mainly carries out the following three aspects of innovation:(1)An intruder screening method based on velocity obstacle method is proposed to ensure that the structure of fuzzy rules does not change with the number of UAVs.Specifically,it includes three steps:collision region partition,collision detection based on velocity obstacle method,and collision time for evaluating the risk of collision(2)A real-time hybrid and fully distributed multi-UAV cooperative collision avoidance method is proposed,which combines the benefits of velocity obstacle and fuzzy control to seek a balance between real-time and security.(3)In this thesis,a velocity optimization method is also proposed.After the fuzzy control generates the candidate velocity,the velocity obstacle method is used to perform the collision check,and the velocity is optimized(modified).