Research On Online Optimization And Decision Algorithm In Dynamic Environment

As an application of mathematics in reality,optimization and decision-making algorithms can solve practically complex(NP-Hard)problems in many societies,such as product distribution route planning,online taxi order allocation,and disaster search and rescue scenarios.However,with the development of society,more and more dynamic optimization scenarios have emerged.In these scenarios,unlike traditional optimization algorithms,the state of the environment is constantly changing with time.Therefore,the performance of traditional optimization algorithms significantly decreases and it is difficult to ensure the optimization effect.Therefore,this thesis selects two scenarios of online order allocation and dynamic scheduling of vehicle distribution in e-commerce and the auxiliary decision-making system for the rescue plan of forest fire heterogeneous robot group,and studies online optimization and decision-making algorithms in a dynamic environment.The main work and innovations include:1.In order to solve the problems of online order allocation and dynamic scheduling of vehicle distribution in e-commerce,this thesis studies the closed-loop execution process of online distribution orders in the field of e-commerce,that is,from online order placement by users to order processing at the distribution center and successful delivery of goods Aiming at this problem,Adaptive Large Neighborhood Search algorithm with Vehicle Balance Strategy(ALNS-VBS)is proposed.This algorithm includes vehicle balancing strategy,vehicle distribution route search,and dynamic task insertion strategy.The vehicle balancing strategy divides the traffic road network into regions,and through the shortage of vehicles in different regions to actively adjust the distribution of vehicles in the road network;In vehicle delivery path search,we designed optimization operators based on space-time sensitivity to solve the problem of separation of orders in time and space;Dynamic task insertion strategy is proposed to solve the problem of receiving new tasks but no idle vehicles in the online distribution process,by dynamically inserting new tasks into the assigned vehicles' paths,the actual task response delay and delivery costs are reduced.Simulation results show that compared with other algorithms,ALNS-VBS can effectively reduce the delivery cost of online tasks and improve the timeliness of online distribution.2.In order to solve the problem of assisted decision-making in the rescue plan of heterogeneous robot groups in forest fires,this thesis designs several targeted forest fire rescue plans based on the fire extinguishing knowledge base,and uses the multi-attribute decision-making method to select the optimal according to the evaluation indexes of forest fire rescue plans Rescue plan.At the same time,this thesis proposes the Artificial Bee Colony with Integrate Firefighting Knowledge(ABC-IFK)algorithm,which aims to assign fire extinguishing tasks in the fire to heterogeneous rescues under the premise of determining the rescue plan The robot,finally,when the rescue robot reaches the target rescue position,it will rescue in the vicinity of the rescue action fuzzy inference machine.This is essentially a greedy strategy.This design is reasonable considering the time loss of the rescue robot moving.In this thesis,20 data sets based on real maps are used for simulation experiments.The results show that the rescue decision aided decision system and ABC-IFK algorithm proposed in this thesis can meet the needs of providing efficient firefighting solutions in forest fire rescue,which can effectively reduce forests.The economic losses caused by the fire,and to a certain extent,reduce the cost of rescue.