Research On Modeling And Optimization Of Mission Planning For Multi-UAVs In Joint Search-and-rescue Mission

Unmanned Aerial Vehicles(UAVs)are widely used in various fields such as reconnaissance,surveillance,search and rescue and other fields on account of the maximum flexibility without the limit of body functions of the crew,the security of performing mission in dangerous areas,the portability without aircrew and the economical efficiency brought by high training expenses.With the wide application of UAVs,the autonomy,intelligent control,and collaborative planning become increasingly important.Especially,more and more scholars have devoted themselves to the study on formation flight,task allocation,path planning,performance assessment and other fields of UAVs due to their abilities to perform missions in boring,bad and dangerous conditions.UAVs have obtained amazing achievements in both military fields and civilian areas.Especially,UAVs are increasingly active at fire monitoring,disaster monitoring,sea rescue,medical supplies delivery and other disaster relief scenes in current days with frequent natural disasters.Now mission planning problems of UAVs are NP-Hard problems involving the cross and integration of numerous disciplines.Many researchers both at home and abroad have made great efforts on control,intelligent algorithms and mathematical models,and proposed many task allocation models and path planning algorithms.For the whole process of mission planning problem,the paper has conducted profitable trails in the fields of threat space modeling,task allocation,path planning and smoothness and mission planning assessment.The main contributions of this paper are summarized as follows:1)The paper has presented a weather threat assessment method based on dynamic discrete Bayesian networks.Considering the application scenarios of search and rescue mission in this paper,terrain and bad weather conditions are the biggest factors of the damage of UAVs.The factors related to weather threat has been abstracted and quantified into four kinds: weather type,the strength of weather impact,relative position and flight category.Also a discrete fuzzy Bayesian weather threat model changing over time has been built and the conditional probability table and the transition probability table have been established according to experts' knowledge and experience.Furthermore,a proper reasoning algorithm has been selected to reason out dynamic weather threat levels.Compared with previous static discrete Bayesian networks and neural networks,this algorithm can obtain more accurate results and reason out precise weather threat level according to threat information before and after when some eigenvalues are missing or wrong.2)The paper has established a task allocation mathematical model and proposed a discrete immune multi-agent algorithm.The task allocation mathematical model is established for the task features of going over the targets as far as possible and high timeliness.And the discrete immune multi-agent algorithm is proposed by learning from the good qualities of artificial immune algorithm and multi-agent system,which focuses on improving algorithm,immune memory operator,neighborhood clone operator,neighborhood suppression operator,neighborhood promotion operator and self-learning algorithm.The algorithm has been successfully applied to static task allocation and dynamic task allocation(three conditions of new added targets,damaged UAV and sudden weather threat).The simulations reveal that the algorithm owns dynamics,adaptability and extensibility.3)The paper has put forward a clonal selection algorithm based on tabu criterion and established dynamic constraints of UAVs about path planning problems.Firstly,the improved A*algorithm has been used to solve out path planning problems.The simulation results show that this algorithm can cut part invalidation nodes and improve solving efficiency,but the solving speed and accuracy are not perfect.Then,taking advantage of the good qualities of the clonal selection algorithm and learning from the tabu search algorithm,the paper has proposed a clonal selection algorithm based on tabu criterion to effectively avoid the problems of low solving efficiency caused by circuitous search during the solving process of clonal selection algorithm.The application on standard TSP test sets of the proposed algorithm has proved its excellent properties.For path planning,the establishment of corresponding encoding methods and the application of prior knowledge for getting vaccines has further improved the convergence rate of the proposed algorithm.For the improved A* algorithm and ant colony algorithm,the proposed algorithm can obtain shorter flight line and faster solving speed.4)The paper has established Bayesian assessment system for mission planning results.The mission planning assessment system based on static Bayesian Networks has been proposed to change previous artificial judging model for task allocation and path planning where the influence elements of mission planning are divided into three kinds: safety,superiority and cooperativity.The relevant quantitative criterions and conditional probability tables are established,and corresponding mission planning levels are reasoned out by variable elimination algorithm.The system can effectively improve the deviations caused by the personal experience of artificial judgment and quickly provide effective evidence for policymakers to choose proper task allocation results and paths.