Abstract:Underwater gliders have been widely used in oceanography for a range of applications. However, unpredictable events like shark strikes or remora attachments can lead to abnormal glider behavior or even loss of the instrument. This paper employs an anomaly detection algorithm to assess operational conditions of underwater gliders in the real-world ocean environment. Prompt alerts are provided to glider pilots upon detecting any anomaly, so that they can take control of the glider to prevent further harm. The detection algorithm is applied to multiple datasets collected in real glider deployments led by the University of Georgia's Skidaway Institute of Oceanography (SkIO) and the University of South Florida (USF). In order to demonstrate the algorithm generality, the experimental evaluation is applied to four glider deployment datasets, each highlighting various anomalies happening in different scenes. Specifically, we utilize high resolution datasets only available post-recovery to perform detailed analysis of the anomaly and compare it with pilot logs. Additionally, we simulate the online detection based on the real-time subsets of data transmitted from the glider at the surfacing events. While the real-time data may not contain as much rich information as the post-recovery one, the online detection is of great importance as it allows glider pilots to monitor potential abnormal conditions in real time.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the abnormal behavior detection problem encountered by underwater gliders during mission execution. Specifically, when underwater gliders operate in the marine environment, they may encounter unpredictable events such as shark attacks and remora attachment, which can cause abnormal glider behavior or instrument loss. To address these issues, this research has developed an anomaly detection algorithm, aiming to monitor the working state of the glider in real - time and alert the operator when any abnormality is detected, so as to take timely measures to prevent further damage.
### Main research contents:
1. **Algorithm development**: An anomaly detection algorithm based on simple data such as glider speed, heading, and trajectory is proposed. This algorithm can be applied to multiple real - world datasets on the basis of theoretical verification and can achieve real - time autonomous operation.
2. **Experimental verification**: Four real - deployment datasets in different scenarios are used to verify the effectiveness and universality of the algorithm. These datasets are from multiple glider deployments at the Skidaway Institute of Oceanography (SkIO) of the University of Georgia and the University of South Florida (USF).
3. **Online detection simulation**: Based on the subset data (SBD data) transmitted in real - time, the online detection process is simulated to achieve real - time interaction with glider operators.
### Research background:
- **Application background**: Underwater gliders are widely used in oceanographic research, such as ocean sampling and surveillance.
- **Challenges**: Due to the complexity of the marine environment and the long - term nature of the mission, gliders may exhibit abnormal behavior due to external interference, resulting in difficulties in positioning and rescue.
- **Deficiencies of existing methods**: Most of the existing anomaly detection methods rely on shore - based manual implementation, cannot achieve real - time detection, and lack real - time experimental verification.
### Solutions:
- **Algorithm framework**: By generating glider speed estimates and flow rate estimates, the algorithm assumes that if the estimated values are within the normal range, there is no anomaly; otherwise, an anomaly may exist. By comparing the flow rate estimated by the algorithm with that estimated by the glider, false alarms can be reduced.
- **Model description**: The glider dynamic model is as follows:
\[
\dot{x}=F_R(x, t)+V_R(t) \Psi_c(t)
\]
where \(F_R\) is the real water flow field, \(x\) is the real position of the glider, \(V_R\) is the real speed of the glider, and \(\Psi_c(t)=[\cos \psi_c(t), \sin \psi_c(t)]^T\) is the real heading angle of the glider.
- **Flow field modeling**: The ocean flow field is modeled using spatio - temporal basis functions:
\[
F_R(x, t)=\theta \phi(x, t)
\]
where \(\theta\) is an unknown parameter, \(\phi = [\phi_1(x, t), \ldots, \phi_N(x, t)]^T\) is a basis function, and its specific form is:
\[
\phi_i(x, t)=\exp\left(-\frac{\|x - c_i\|^2}{2\sigma_i^2}\right) \cos(\omega_i t+\upsilon_i)
\]
### Experimental results:
- **Large - scale experiments**: Through the analysis of four different deployment datasets, the effectiveness and universality of the algorithm are verified. These datasets include the deployment data of the Franklin, USF - Sam, USF - Gansett, and USF - Stella.
- **Online detection**: Based on the real - time transmitted SBD data, the online detection process is simulated, and the results show that the algorithm can effectively detect anomalies in real - time data.
### Conclusion:
The anomaly detection algorithm proposed in this research performs well on multiple real - deployment datasets, can achieve real - time monitoring and anomaly detection, and is helpful to improve the safety and success rate of underwater glider missions. Future work will further improve the estimation accuracy, construct a data - driven framework by combining a large amount of glider data, and consider the impact of anomalies on flow rate estimation.