SCIENCE CHINA Information Sciences, Volume 63 , Issue 7 : 170209(2020) https://doi.org/10.1007/s11432-019-2751-4

Data fusion using Bayesian theory and reinforcement learning method

More info
  • ReceivedAug 30, 2019
  • AcceptedNov 29, 2019
  • PublishedApr 30, 2020


There is no abstract available for this article.


This work was supported by Major Projects for Science and Technology Innovation 2030 (Grant No. 2018AA0100800) and Equipment Pre-research Foundation of Laboratory (Grant No. 61425040104).


[1] Du Y K, Jeon M. Data fusion of radar and image measurements for multi-object tracking via Kalman filtering. Inf Sci, 2014, 278: 641-652 CrossRef Google Scholar

[2] Nguyen H, Cressie N, Braverman A. Spatial statistical data fusion for remote sensing applications. J Am Statistical Association, 2012, 107: 1004-1018 CrossRef Google Scholar

[3] Bass T. Intrusion detection systems and multisensor data fusion. Commun ACM, 2000, 43: 99-105 CrossRef Google Scholar

[4] Wan S P. Method of fusion for multi-sensor data based on fisher information. Chin J Sensor Actuat, 2008, 21: 2035-2038. Google Scholar

[5] Li X X, Peng Z H, Liang L. Policy iteration based Q-learning for linear nonzero-sum quadratic differential games. Sci China Inf Sci, 2019, 62: 52204 CrossRef Google Scholar

[6] Zhang T, Huang M, Zhao L. Learning structured representation for text classification via reinforcement learning. In: Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence. 20181-8. Google Scholar

[7] Yan X H, Zhu J H, Kuang M C. Missile aerodynamic design using reinforcement learning and transfer learning. Sci China Inf Sci, 2018, 61: 119204 CrossRef Google Scholar

[8] Wang H H, Wu Y, Fu Y. Data fusion using empirical likelihood. Open J Statist, 2012, 2: 547-556 CrossRef Google Scholar

[9] Wang W, Zhou J, Wang R. A method of the multi-sensor data fusion. J Transduc Technol, 2003, 22: 39-41. Google Scholar

  • Figure 1

    (Color online) (a) Error histogram; (b) the error curve of random samples.


    Algorithm 1 Reinforcement learning based Bayesian data fusion algorithm

    Require:The observations $O_1,~O_2,~\ldots,~O_m$, the variances of sensors $\sigma_1,~\sigma_2,~\ldots,~\sigma_m$.

    Output:The fused data.

    Initialize $Q=0$, and set the discount factor $\gamma$;

    for each episode

    for $t=1$ to $m$

    while state $s_t$ is not terminal do

    Initialize state $s_t$;

    $a'\leftarrow$ action in state $s_t$;

    Take action $a'$, calculate reward $r$, and obtain the next available state $s'$;

    Update $Q$ according to (5);

    Calculate the fused data ${\hat~{O}_t}$ based on the Bayesian theory (3);

    $s_t\leftarrow$ optimal new state $s'$;

    end while

    end for

    end for