Research On Trajectory Prediction Of Nodes In WBAN

With improvement in people’s standards of living，Wireless Body Area Network(WBAN),an important branch of Internet of things, shows vast development spaceand broad market prospects. At present, many researches on WBAN are still at theprimary stage. Prediction of motion trajectory of nodes is the basis of data fusion,routing control and especially power control in WBAN. It has important significancefor prolonging survival time and keeping connectivity for WBAN, and has greatresearch value.Pointing to prediction of motion trajectory of nodes in WBAN, proposed apredicting method based on grey model. This method is suitable for the applicationenvironment on aspects of accuracy, real-time performance and computationalcomplexity.When grey prediction is applied to shock disturbed system, there is aninconsistency between quantitative prediction and the conclusion about qualitativeanalysis. To solve this problem, we proposed a kind of buffer operators with variableweights under axioms of buffer operator. The adjustment degree of buffer operatorschanges with the variable weights. The buffer operator proposed in this paper isweakening or strengthening buffer operator with different value range of the variableweights. The adjustment degree of the buffer operator varies in the same directionwith the variable weight when the operator is a weakening one, but the adjustmentdegree varies in the other direction when it is a strengthening operator. Theseconclusions help us adjust the weight reasonably.Then, according to the character of the buffer operator, we added a self-adaptivestrategy of weights adjustment to the prediction procedure, which enable us to weakenor strengthen the observed series by taking different values of the weight. The carrier of prediction in WBAN is node itself. In previous studies, the weights are appointedby the genetic algorithm or artificial experience. Considering the low computingcapability, small storage space and difficulties to monitoring, the two ways above canhardly be applied to WBAN. So we proposed this strategy. It can self-adaptivelychange the weight based on the trend of the input data, so that the adjustment degreechanges self-adaptively. The predicting model can automatic adjust without artificialmonitor when this strategy is applied. This strategy improved the accuracy of themodel with little amount of calculation.Our experiments are based on the MSR Daily Activity3D dataset provided byMicrosoft. Data in this dataset is representative human daily activity trajectoriescollected by Microsoft Kinect. On this data, the predicting performance improvesusing the buffer operators when compared with the traditional grey model. Then thepredicting performances get another improvement when we added in the proposedstrategy.