4results about How to "Realize multi-objective optimization" patented technology

The invention relates to the technical field of ocean platform pipeline design, in particular to an ocean platform pipeline laying method based on reinforcement learning, and the method comprises the following steps: S1, dispersing a pipeline laying region into a three-dimensional grid matrix according to the physical size of an actual cabin of an ocean platform, and marking a pipeline starting point, a pipeline ending point and an impassable region; s2, performing Q-Learning algorithm parameter initialization configuration, defining an action space adaptive to the linear movement characteristics of the ocean platform pipeline, constructing a Q value matrix adaptive to three-dimensional space coordinates and actions, and performing initialization; s3, entering a training round, and continuously updating the value evaluation matrix by dynamically adjusting a greedy criterion, a multi-dimensional reward mechanism and a time sequence difference learning rule; and S4, after the training is completed, starting from the starting point based on the converged Q value matrix, selecting an optimal action through a greedy to generate a final pipeline path, and improving the quality and search efficiency of pipeline laying on the ocean platform.

The invention provides a multi-target optimization method for multi-DNN model division and request scheduling for an intelligent internet of things. The method comprises the following steps: S1, initializing a multi-target reinforcement learning model parameter and a genetic algorithm population parameter; s2, collecting current state characteristics of the environment, and generating a multi-DNN model division decision by using a multi-target reinforcement learning model based on multi-task learning; s3, calling a genetic algorithm to search an optimal request scheduling scheme, and calculating a multi-target reward; s4, performing parameter updating on the multi-target reinforcement learning model by adopting a near-end strategy optimization algorithm; s5, judging whether reinforcement learning reaches a preset number of iterations or not, if yes, going to S6, and if not, continuing S2; and S6, deploying the trained model to intelligent Internet of Things center equipment, obtaining environment characteristics in real time and outputting an optimal decision, so as to control Internet of Things terminal equipment to cooperatively complete multiple DNN reasoning requests. According to the method, the advantages of multi-objective reinforcement learning and the genetic algorithm are combined, and multi-objective optimization of system delay and energy consumption is achieved.

The invention discloses a microwave heating temperature control method and system based on a self-adaptive whale optimization algorithm. The method comprises the following steps: S1, constructing a multi-objective optimization problem of microwave heating temperature control; s2, based on the feasible region of the control variable, randomly generating an initial whale population containing a plurality of individuals; s3, performing simulation evaluation on each individual in the initial whale population; controller parameters corresponding to the individuals are input into the microwave heating system simulation model, temperature response data of the system are obtained, and the value of a comprehensive fitness function J is calculated according to the temperature response data; s4, executing self-adaptive whale optimization algorithm iteration, and updating the whale population; and S5, when an iteration termination condition is satisfied, outputting a global optimal individual, and applying a controller parameter corresponding to the global optimal individual to an actual microwave heating control system. The method has the advantages of adaptively improving temperature uniformity, response speed, control precision and the like.

This invention discloses a dynamic spectrum allocation and multi-path collaborative transmission communication method and system for unmanned platforms. The method involves: first, real-time scanning of the operating frequency band, synchronously collecting signal strength and interference types of the target frequency band, generating a spectrum occupancy heatmap, and classifying interference signals; then, generating an availability probability map of each sub-channel within a future time window to predict the spatiotemporal distribution of interference sources; next, constructing a deep reinforcement learning (DRL) model to achieve dynamic allocation of transmission resources; finally, when the detected interference intensity exceeds a preset threshold, activating a multi-band, multi-path concurrent transmission and network coding mechanism, with the main path using a low-frequency band to ensure coverage, and the backup path using a high-frequency band to ensure rate. The receiving end processes the received data using a confidence-weighted fusion algorithm. The system includes a spectrum sensing module, a decision engine, and a multi-path transmission module. This invention improves communication efficiency and anti-interference performance, is highly adaptable, flexible in deployment, and has good scalability.