Paper
Bias Compensation Estimation in Multi-UAV Formation and Anomaly Detection
- Authors:
- Jian-hong Wang; Yan-xiang Wang; Yong-hong Zhu
- Abstract
- A data-driven method is used to detect anomalies in the formation of multi-UAVs from input and output data sequences corresponding to each UAV. As this special method does not require prior structural information of the UAV, it avoids the complexity introduced by multi-hypothesis testing and probability inequalities. The established nonlinear unknown form corresponding to each UAV can be well approximated by basis functions and converted to a linear regression form which includes one multiplication operation between a regression and a parameter vector. The parameter vector is identified by the least square method; one anomaly detector is then constructed based on the derived residuals. When only the inputs are included in this nonlinear form, the anomaly detector using the consistent parameter estimations demonstrates improved performance. However, when the inputs and outputs are all included simultaneously, the parameter estimations derived by the least square method are all biased. These biased estimations seriously affect the final anomaly detector. In order to avoid this bias, the bias compensated terms are added into the bias estimations. Furthermore, this paper also proves the specified expressions representing the derived bias compensated terms and the consistent property. A method to replace some of the matrices in the bias compensated terms is also proposed. Finally, results of an example simulation confirm the theoretical identification results.
- Keywords
- Multi-UAV Formation; Anomaly Detection; Nonlinear System Identification; Bias Compensated Estimation
- StartPage
- 40
- EndPage
- 50
- Doi
- 10.18005/JCSE0401005