Parameter identification based on genetic algorithm for PI controller in wireless sensor networks

Wireless sensor networks together with hundreds of thousands of cheap micro-sensor nodes allow users to collect and process information in diverse environments. However, each node has limited resources to storage or compute due to its unique features. Therefore, for the regulation of controllable traffic in wireless sensor networks, optimization of control parameters has a direct influence on the proportional-integral (PI) congestion controller. A novel approach by using genetic algorithm (GA) to parameter identification for PI controller in the intermediate switch node is proposed in this paper. In GA evolution process, binary system matrix encoding. technique is adopted, and the optimal feedback coefficients cm be obtained to make objective function minimum in a large space. Our preliminary simulation results show that the optimal PI controller gains are scheduled in response time and dynamic buffer occupancy.