Research On Navigation And Obstacle Avoidance Of Unmanned Surface Vehicle In Complicated Marine Environment For Defense Mission

With the development of modern warfare and the advance of science technology,the unmanned weapon system is developing towards automation,systematization and intellectualization.It has broad application potential in both military and civilian fields.The modern unmanned systems such as unmanned surface vehicles(USVs),unmanned aerial vehicles(UAVs),unmanned ground vehicles(UGVs)and unmanned underwater vehicles(UUVs),that can monitor in a wide range of domains and respond quickly to emergencies.The unmanned surface vehicles have become the developed trend of marine military equipment construction.Its flexibility and intelligence make it widely used,especially in the field of dangerous surface,which extends its detection range and operation capability to the greatest extent.Therefore,in the field of defense research of USV,its intelligence and autonomy are the key.In this paper,we analyzed the characteristics of USV,such as the autonomous navigation obstacle avoidance and intelligent task assignment capability.And we research the defense capability of USV based on group intelligence and deep reinforcement learning,which is considered as an important approach to artificial intelligence.Therefore,the research work of this paper is summarized as follows:Firstly.The construction of defense scenario in sophisticated marine environment.The basis to solve the USV defense problem is setting the environment scene,so we construct the multi-USV defense scenarios and build the corresponding models.In order to ensure the maximum coverage and real-time perception of monitoring area,the multiple USVs need to keep the corresponding formation and distribute around the target.There are two kinds of USV: master USV and sub-USV.and they monitor the area along the patrol line in daily situations.When suspicious invaders appear,the master USV collects the sensing information for all subUSVs,and carries out information fusion.After decision-making,it issues defense instructions to the corresponding sub-USVs.Meanwhile,each sub-USV perform it interception missions according to the received instructions,and feedbacks the real-time state information.The collaboration and cooperation among multiple USVs can resist the invasion of suspicious boats in many directions,and effectively apply tactics to enhance their own security.At the same time,aiming at the unique and effective control of master USV in the defense system,this paper puts forward the coordinated control technology,using the idea of master-slave,to solve the problem.when problems arise in the defense system,the master USV can be replaced by other USV.Preventing the defense system from falling into chaos and ensure its continuous and effective operation.Secondly,the multiple USVs defense based on group intelligence.The collaboration among multiple USVs to accomplish defense missions together,which can show the advantages of group operations.The collaborative way can greatly improve the intelligence of individual behavior,and accomplish complex missions that individuals cannot fulfill.In this paper,inspired by the group behavior in nature,such as ant foraging,bee nesting,wild goose flying southward and so on,we consider that multiple USVs show coordinated movement behavior and cooperate with each other to resist external threats.Thus,we proposed dynamic overlay reconnaissance algorithm based on genetic idea(GI-DORA)to solve the problem of multi-UAV multi-station reconnaissance.Moreover,we developed continuous particle swarm optimization based on obstacle dimension(OD-CPSO)to optimize defense path of USVs to intercept intruders.In addition,under the designed defense constraints,we proposed dispersed particle swarm optimization based on mutation and crossover(MC-DPSO)and real-time batch assignment algorithm in emergency(RTBA)for formulating combat defense mission assignment strategy in different scenarios.Although the behavior of single USV is simple and its capability is limited,they can emerge very complex collective intelligence features when they work together.Thirdly,the autonomous navigation and obstacle avoidance algorithm based on deep reinforcement learning is proposed.The ability of autonomous navigation and obstacle avoidance is indispensable part when performing various missions in sophisticated marine environment.It is also an important manifestation of its intelligence.With the development of modern theory and technology,especially deep learning and reinforcement learning,which bring vitality to the development of autonomous control.In this paper,we propose autonomous navigation and obstacle avoidance algorithm based on reinforcement learning(ANOA).The problem is modeled by Markov decision process,and then the simulation platform of navigation is constructed.The deep reinforcement learning algorithm is used to guide the intelligent USV to approach destination gradually and without collision.Furthermore,introducing the environmental uncertainty while taking into consideration dynamic characteristics of obstacles,we illustrated its performance of autonomous navigation and obstacle avoidance for USVs under strategic guidance of proposed algorithm.In addition,we designed corresponding reward functions as the unique feedback of navigation ability,which reflect the fitting quality of deep neural network on the aspect of autonomous navigation.Meanwhile,relevant experiments are carried out based on the constructed unmanned surface vehicles simulation platform,and simulation results verified the rationality and effectiveness of proposed algorithm.