How to calculate the key technologies of WSNs energy consumption optimization through population intelligence and genetic algorithms?
Swarm Intelligence (SI) and genetic algorithms (Genetic Algo⁃rithm, GA) both fall under the scope of research in the field of artificial intelligence.
SI is a general term for the collective intelligence behavior of a class of decentralized self-organizing systems, that is, based on the aggregation of individual group members to show independent intelligence, which is suitable for solving optimization problems and helps to achieve NP problems that cannot be solved by polynomial time complexity. GA is a kind of self-organizing, adaptive artificial intelligence technology that simulates the process and mechanism of biological evolution in nature to solve optimal problems.
After forming the initial population through coding, the task of genetic manipulation is to impose certain operations on individuals according to the environmental fitness of individuals in the population, so as to realize the evolutionary process of survival of the fittest.
Many WSNs applications have a large number of widely distributed sensor nodes that are battery-powered and difficult to replenish, with severe energy limitations. Reducing and balancing node energy consumption and maximizing network lifetime have become the primary design goals of WSNs.
Node energy consumption is closely related to WSNs network protocols, and well-designed routing algorithms can greatly reduce and balance node energy consumption and extend network lifetime. It is of great significance to use natural element heuristic intelligent algorithms such as SI and GA to realize the WSNs routing protocol to comprehensively coordinate and schedule data communication and computing processing of various on-network sensor nodes, and optimize network energy consumption.
First, the key technology of WSNs energy consumption optimization
The WSNs clustering protocol comprehensively considers the fixed energy consumption of nodes sending and receiving data and the energy consumption of the wireless power amplifier of the transmitter that changes with the transmission distance, and divides the network into several local clusters.
The cluster head nodes in each cluster consume additional energy and are responsible for receiving cluster member data and sending it to the base station. The clustering structure makes the overall network scalability good, and has better performance than the direct communication between nodes and base stations and the minimum transmission energy routing between nodes, which has become the research focus of WSNs routing algorithms.
The two core problems of clustering protocols are:
(1) formation of clusters;
(2) Establish a multi-hop forwarding path for data between the cluster head and the base station.
The communication range of the base station in the small area monitoring environment covers the entire monitoring area, and the cluster head and base station do not need to forward data through the relay node, only need to consider "forming a cluster", and then the cluster head and the base station single-hop communication; The communication range of base stations in medium-area and large-area monitoring environments cannot cover the entire monitoring area, and the cluster head and base station forward data through relay nodes, which requires comprehensive consideration of the above two problems.
Second, establish the mathematical model and energy consumption model of WSNs based on the residual energy index of the node
The first-order radio model is a general energy consumption model of WSNs, within which the energy consumption formula of data sending, receiving, fusion, calculation, and perception is defined to simulate the real energy consumption of the network. WSNs are divided into homogeneous WSNs with equal initial energy and heterogeneous WSNs with different initial energy of nodes.
With the in-depth analysis of the remaining energy of nodes and round-by-turn clustering, various WSNs mathematical models are mapped to multi-level heterogeneous models.
WSNs nodes are usually randomly scattered in the monitoring area and stationary, and the node location, node degree, distance from other nodes, distance from the base station and other indicators are randomly distributed.
Although the initial energy of homogeneous WSNs nodes is equal, and the network protocol used also intends to keep the energy consumption of each node consistent in each round, the actual energy consumption of each node cannot be the same due to the difference in the above indicators.
Therefore, isomorphic WSNs are essentially special cases of heterogeneous WSNs. In the process of establishing the network protocol, as long as the remaining energy of the node is fully considered, the homogeneous and heterogeneous WSNs can be unified in the multi-level energy heterogeneous WSNs mathematical model.
1. Single hop path between clusters
The optimal number of cluster heads in this scenario is direct communication between the member nodes and the cluster heads, cluster heads and base stations, the number of link hops is 2, and the network topology is relatively fixed, as shown in the following figure.
The factors affecting the calculation formula of the optimal number of cluster heads in the scenario of intercluster single-hop routing include the position of the base station relative to the monitoring area, the main and auxiliary cluster head mechanism, the control package, the free space model and the multi-path attenuation model, etc., and comprehensively use mathematical tools and concepts such as calculus, polar coordinates, and probability distribution to calculate the optimal number of cluster heads combined by 16 factors and the mathematical formula of the total energy consumption of the network in one round, as shown in the following table.
2. Single hop path between clusters
The single-hop routing scenario between clusters only needs to consider the clustering process, and a particle of the natural element heuristic intelligent algorithm searches for a solution to the optimization problem, which is mapped to the collection of all cluster heads in the monitoring area, and the quality of the solution is evaluated by the multi-objective fitness function.
The three SI algorithms of PSO, ABC and FA respectively refer to the nectar source locations (hereinafter collectively referred to as particles) corresponding to particles, fireflies and bees, and find better solutions through mutual cooperation, learning and attraction between particles. The above process must establish a mapping relationship between a continuous amount of particles and a discrete cluster head set, and when the position of the solution represented by the particle moves in continuous space, it is necessary to find the corresponding discretized cluster head set according to certain rules. The solution of the GA algorithm is called chromosome, and it continues to evolve through chromosome selection, crossing, and mutation operations, so as to search the solution space to find a better solution.
The CLON⁃ALG algorithm simulates the human immune mechanism, introduces antibody selection, replication, cloning value-added hypervariation, mutation and other operations to retain and search for high-quality antibodies, and quickly find better solutions. In the process of algorithm implementation, how to determine the Pareto multi-objective fitness function and its weight coefficient is a difficult point, which has a great impact on the performance of the algorithm, and it is also necessary to consider the influence of cluster head and base station distance on cluster scale.
3. Multiple hops between clusters
Because the number of hops between the cluster head and the base station cannot be determined in advance, it is not possible to mathematically derive the optimal number of cluster heads formally. In order to generate a dynamic number of preferred cluster head sets, it is not advisable to include all nodes in the cluster head candidate set at chromosome or antibody initialization for computation greatly increases the computational load. According to certain filtering rules, the cluster head candidate node set can be conditionally determined, and the number of collection elements is much smaller than the total number of nodes.
