Improving the Accuracy of Vessel Traffic Monitoring System Displays Essay Example

use of Kalman and Particle Filters.

The text discusses how information from a database at UCSC containing AIS information of ships in Colombo-South Harbour is collected. The survey focuses on the ship's location, velocity over land (SOG), class over land (COG), and true heading. A ship is selected from the database to implement the Kalman and Particle Filters. The text highlights an issue with the original information, pointing out that there are noticeable variations in AIS measurements. This makes it difficult to track the ships continuously in the VTMS.

The Kalman Filter relies on both Kalman addition and the anticipation clip as crucial elements. An analysis is performed to determine the impact of these factors. The results of the analysis demonstrate that the Kalman Gain reaches its peak when there is higher uncertainty in the process or model compared to uncertainty in the measurements. Additionally, it becomes apparent that the Kalman Gain stabilizes quickly over time. When Q is significantly large, the estimations precisely track the measurements. The filter's anticipation clip is chosen as the median value of the time difference between signals.

Then the Kalman Filter is used on the AIS information. The Kalman Filter is able to make predictions when AIS signals are lost. Additionally, there are several consequences including:

• Some values predicted by the Kalman Filter deviate from the actual measurements initially, but as time progresses, the predictions converge to the actual measurements.
• The deviations may be due to the assumption of linearity in a nonlinear model.

6.3 Discussion of Verification and Validation of Kalman Filter

The Kalman Filter is then applied to a simulated dataset created by removing some records from the original dataset. It is clear that the Kalman Filter predictions agree with actual predictions in some periods, but deviate in others. This may be attributed to a lack of records in the simulated dataset. To verify the previously developed Kalman Filtered estimates using the simulated dataset, the original dataset is used as a reference.

The confirmation and proof procedure focuses on checking the proper functionality of the Kalman Filter in improving the AIS data. The results indicate that the estimation is much smoother compared to when only artificial dataset is added. However, there are still instances where appraisals deviated from the actual information. This could be due to natural discontinuity present in the AIS data. The residual analysis, conducted to examine the operation of the Kalman Filter, shows that the residuals appear to scatter around zero randomly, indicating the proper functionality of the Kalman Filter.

6.4


Discussion of the


Atom Filter

Implementing the Particle Filter is a tedious project.

An attempt is made to develop the Particle Filter to enhance AIS information. The normal distribution is chosen as the importance density. Atoms represent possible ship locations in the following measurement. In the study, higher weights are given to atoms that are closest to the ship compared to those that are farther away. The findings are as follows:

• When using the Particle Filter to improve accuracy of AIS data, there are significant differences between Particle Filtered estimates and actual measurements.
• However, there are a few instances where Particle Filtered estimates agree with actual data.

6.5 Conclusion

The research project focuses on addressing the use of

Kalman and Particle Filters to improve accuracy of AIS-based position data.

• The Kalman Filter is more accurate when dealing with missing measurements. Deviations between estimates and actual data may occur due to the assumption of linearity in a nonlinear system.
• The Particle Filter is a suitable choice when the linearity assumption is violated. It can provide estimates for missing information, but its effectiveness heavily relies on selecting the appropriate importance density.

6.6 Restrictions of the Study

The study faced a few limitations and challenges that should be considered for future research.

• When developing the Kalman Filter, we had to input the procedure noise covariance and the measuring noise covariance matrix. However, it was not possible to give a straightforward value for these inputs and they had to be decided intuitively.
• In order to use the Kalman Filter, we needed to provide a reference information set. Typically, this information comes from another source such as radar data or AIS input. Unfortunately, we were unable to obtain that information for our research. As a result, we had to use the same dataset for both instances, which limited the satisfaction of the obtained results. It would have been more satisfactory if we could have used a different dataset from a different source.
• Implementing the Particle Filter was a challenging and time-consuming task. Due to time restrictions, we could only use the normal density as the importance density in the Particle Filter. It would have been better to try a different distribution as the importance distribution.
• When assigning weights in the Particle
  

Filter, we did not consider the existence of land locations. This proved to be a tedious project and could have improved accuracy of results if it had been considered in research.

• A more appropriate method for assigning weights in research would involve considering atoms closest to ship with higher weights compared to those further away.

The current research does not account for the fact that atoms within a certain distance of the ship will have more weight, regardless of their proximity to the ship's current speed (e.g.5 kmph).

6.7 Further Research:
Future studies in this area could consider the following:

- Due to time and other constraints, implementing the Particle Filter was challenging.To improve estimation in Particle Filtering, different importance distributions could be used.Additionally, developing relevant Matlab codes to enhance weight allocation to atoms would be beneficial.
- The process of weighting atoms could be further improved by incorporating the ship's speed into weight assignment.Additionally, the research did not consider land locations, which could have improved the accuracy of the results.Including these factors in future studies could enhance findings.