Research On Mission Planning Approach For Computational Agriculture Based On Multi-Objective Optimization

As the demand for food is growing continuously and science technologies are developing rapidly,the traditional agriculture has turned into computer-based smart agriculture.The intelligent agriculture is capable of performing smart operations,such as producing and controlling,by perceiving,computing and analyzing the agricultural data.Intelligent agriculture not only decreases the human labor and financial investment,but also allocates the limited resources reasonably.Meanwhile,the intelligent agriculture is able to produce more agricultural products and increase the equality of products in an environmental-friendly way.The core of intelligent agriculture is computation and intelligence.Intelligent producing and controlling rely on analyzing and processing agricultural data.The intelligent agriculture is capable of providing theoretical foundations and precise suggestions for agricultural processes.This paper proposes a mission planning approach based on computation agriculture and multi-agent system theory.Complex agricultural missions are assigned to agricultural-operation actors reasonably by following generated strategies.The mission planning strategies are computed by the multi-objective optimization algorithm.Lastly,the proposed approach is verified through simulation.The research objectives of this paper is concluded as follow.(1)Research on the theory of multi-agent system.This research focuses on analyzing and summarizing the definition of agents,basic model of agents,definition of multi-agent system,characteristics and general architectures of multi-agent system.Based on the research,components of computational agriculture system are treated as agents and these agents could construct a computational agriculture system,as a multi-agent system.(2)Research on multi-objective optimization problem.The mission planning problem in computation agriculture is treated as a multi-objective optimization problem.Multiple objective functions are proposed,including beneficial and cost objective functions.The beneficial objective function considers the expected benefit,probability of handling the tasks and efficiency of agricultural machineries.The cost objective function considers the energy consumption and equipment loss of agricultural machineries during the mission execution.(3)Research on multi-objective optimization algorithm.This research analyzes and summarizes two multi-objective optimization algorithms,including particle swarm optimization and genetic algorithm.In order to generate the optimal mission planning strategy,an improved optimization algorithm,MP-PSOGA,is proposed.The main body of proposed algorithm is particle swarm optimization,embedded with crossover and mutation operators from genetic algorithm.Meanwhile,the parameters of proposed algorithm have a dynamic updating mechanism,for the purpose of increasing the convergence speed.The dynamic parameters are applicable to the unpredictable changes of agricultural system model.(4)Research on mission planning strategy.Firstly,the establish mission is decomposed to several sub-missions,called tasks.The tasks are sorted by the priority.During the mission planning process,a negotiation mechanism is introduced.The negotiation mechanism is implemented by holding task auctions.After placing the bids,the auctioneer calculates the comprehensive benefit and assigns each task to the most appropriate agents.Meanwhile,this paper considers the dynamic factors,which are easily disturbing agricultural machineries to follow the original strategy.Hence,a mission re-planning mechanism is designed in order to avoid such situations.(5)Simulation and result analysis.A agricultural mission called pesticide spraying is assigned to several agricultural machineries in order to verify the feasibility and usefulness of proposed mission planning approach.