Bio-Inspired Based Real-Time Task Assignment Algorithm for Heterogeneous Multiprocessor Systems

In this modern era, task assignment is a serious concern due to the growth of innovative real time smart gadgets and increasing needs of data management. Many researchers have been introduced machine learning strategies to provide some resolution to the task assignment issues and dynamic processor selection based on the tasks. In general cases, some pre-defined datasets have been used to process the task and the processor selection which is based on the parameter utilization and high-priority ranking scenarios. However, in some cases the needs are different and the expectation of processor selection based on parameters such as power, speed, energy. The dataset has to be created manually according to the needs. The dataset generation process is based on two categories, one is manually generating the dataset for training purposes and the other one is dynamically storing the real-time values from the devices & a tasks are arranged based on some attribute values like utilization. These tasks are considered to be the training sets and some of the present coming tasks are taken as a testing data used for processor selection. In this work, an efficient new hybrid task assignment algorithm is proposed to provide efficient task assignment solution for the heterogeneous environment. The objective of the proposed work is to classify the tasks and allocate them to exact processors based on the processor utilization factor and the energy constraints. The proposed algorithm has been developed with the integration of Squirrel Optimization Algorithm (SOA) and Local Search Algorithm. The Support Vector Machine Classifier based on the Radial-Basis-Function (RBF) has been used to improve the performance of the proposed algorithm. From the results, it has been observed that the proposed algorithm provides enhanced accuracy and training accuracy in terms of task assignment through the ROC curve.