TY - JOUR
T1 - An efficient statistical strategy to monitor a robot swarm
AU - Harrou, Fouzi
AU - Khaldi, Belkacem
AU - Sun, Ying
AU - Cherif, Foudil
N1 - KAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): OSR-2019-CRG7-3800
Acknowledgements: This Publication is based upon work supported by King Abduallah University of Science and Technology(KAUST) Office of Sponsered Resaerch (OSR) under award No: OSR-2019-CRG7-3800
PY - 2019/10/31
Y1 - 2019/10/31
N2 - Detecting anomalies in a robot swarm play a core role in keeping the desired performance, and meeting requirements and specifications. This letter deals with the problem of detecting anomalies in a robot swarm. In this regards, an unsupervised monitoring approach based on principal component analysis and k-nearest neighbor is proposed. The principal component analysis model is employed to generate residuals for anomaly detection. Then, the residuals are examined by computing the proposed exponentially smoothed k-nearest neighbor statistic for the purpose of anomaly detection. Here, instead of using parametric thresholds derived based on the Gaussian distribution, a nonparametric decision threshold is computed using the kernel density estimation method. This provides more flexibility to the proposed detector by relaxing assumption on the distribution underlying the data. Tests on data from ARGoS simulator show efficient performance of the proposed mechanism in monitoring a robot swarm.
AB - Detecting anomalies in a robot swarm play a core role in keeping the desired performance, and meeting requirements and specifications. This letter deals with the problem of detecting anomalies in a robot swarm. In this regards, an unsupervised monitoring approach based on principal component analysis and k-nearest neighbor is proposed. The principal component analysis model is employed to generate residuals for anomaly detection. Then, the residuals are examined by computing the proposed exponentially smoothed k-nearest neighbor statistic for the purpose of anomaly detection. Here, instead of using parametric thresholds derived based on the Gaussian distribution, a nonparametric decision threshold is computed using the kernel density estimation method. This provides more flexibility to the proposed detector by relaxing assumption on the distribution underlying the data. Tests on data from ARGoS simulator show efficient performance of the proposed mechanism in monitoring a robot swarm.
UR - http://hdl.handle.net/10754/659969
UR - https://ieeexplore.ieee.org/document/8889388/
UR - http://www.scopus.com/inward/record.url?scp=85078813582&partnerID=8YFLogxK
U2 - 10.1109/JSEN.2019.2950695
DO - 10.1109/JSEN.2019.2950695
M3 - Article
SN - 1530-437X
VL - 20
SP - 2214
EP - 2223
JO - IEEE Sensors Journal
JF - IEEE Sensors Journal
IS - 4
ER -