Research On Key Technologies Of Sensor Cloud

In recent years,wireless sensor networks(Wireless Sensor Networks,WSNs)have been widely used in medical,military,and environmental infrastructure fields.However,the constraints of wireless sensor networks in terms of computing,storage,and energy have greatly restricted their application.The emergence and development of cloud computing has greatly expanded the application of wireless sensor networks.As a combination of wireless sensor network and cloud computing,the sensor-cloud system has the advantages of strong real-time perception,fast calculation speed and large storage space.However,due to problems such as frequent data requests,low task success rate,and insufficient node energy level,the existing research on sensor-cloud systems is faced with great challenges.In order to overcome this challenge,this thesis conducts a comprehensive research on the sensor-cloud system from the perspectives of data prediction,resource scheduling and task offloading.First of all,the prediction of sensor-cloud data was studied in this thesis.In the sensor-cloud system,data prediction technology can effectively reduce energy consumption and time delay.The simulated annealing algorithm to optimize the state transition probability criterion of the BP(Back Propagation,BP)neural network algorithm was used in this thesis.The peak recognition mechanism was used in the BP neural network to improve the accuracy of peak prediction.The impact of cloud system energy consumption and latency was studied.The experimental results show that,compared with other schemes,the improved BP neural network algorithm has the best prediction performance and can effectively reduce energy consumption and delay of the sensorcloud system.Secondly,the resource scheduling of sensor-cloud was studied in this thesis.Aiming at the problem of low task success rate in existing virtual machine group scheduling algorithms,an Improved Algorithm for Scheduling Task Based on Priority(IMPBA)was proporsed in this thesis.Different scheduling schemes according to the size of the task priority gap was used in this algorithm.Experimental results show that compared with other schemes,IMPBA algorithm can effectively improve resource utilization and task success rate,and reduce task processing time.Finally,the offloading of sensor-cloud tasks was studied in this thesis.Aiming at the problem of insufficient node energy in the sensor-cloud system,a fog computing-based PADO(Predictive and Dynamic Offloading,PADO)algorithm was proposed in this thesis.The dynamic unloading problem was transformed into a random optimization problem.The total power consumption of the system under the premise of ensuring the stability of the system was minimized in the prposed algorithm.Experimental results show that compared with other solutions,the system power consumption and the amount of task backlog were reduced in PADO algorithm.