Research On The Problem Of Decision Fusion In The CPS

Cyber-Physical Systems (CPS) is a new intelligent complex system that generates and processes large amounts of data. To improve the ability of in-formation abstraction, data fusion is usually as a focus to meet the application requirements in CPS. Considering some characteristics of CPS are different from the existing system's such as close loop feedback/auto-control in a long-term period and the interference like human factors caused by the openness of CPS, this paper's research focus on the problem of decision fusion in the CPS.The current research situations are analyzed based on the research of CPS and data fusion, relevant theoretical backgrounds are presented and decision level fusion algorithms of the CPS are put forward with the key analysis.On the one hand, a novel decision fusion algorithm based on time series (TS-DLF) is proposed in view of the long-term valuable historical data of the CPS. The weight for each time series undergo decay on a timescale and impact factor that we set can be adjusted by internal feedback for the improvement of the accuracy of fusion. Moreover, simulations are performed on mobile medical platform to validate the algorithm and the results show that the historical data have the ability to influence the decision fusion for making an overall judgment and the system can achieve a stable state.On the other hand, since the interference caused by human factors always existed in CPS and CPS has been more and more widely used, a decision and prediction model based on machine learning (ML-DPM) for CPS is designed. The systems can learn complicated knowledge through training Deep Belief Network (DBN) and adopting Contrastive Divergence(CD) algorithm. Simulations are performed in view of four different samples of the model. The simulation results show that the network can solve the problem of the feature extraction and relational mapping that easily affected by human factors-the network effectively achieves decision fusion and prediction in CPS.