In this section, we offer some conclusions and some suggestions regarding this research.
Section 5.1 and 5.2 discuss conclusions and suggestions, respectively.
5.1 The Conclusions
We draw some conclusions about our model as follows:
1. According to the testing results, our model is suitable in general. Entropies represent the uncertainties of sensors. There is no requirement that these sensors must have the same amount of data.
2. Our model can be used to reflect sudden changes of speeds. The changes of traffic conditions are considered continuance. However, a sudden change in speed during a short period of time sometimes occurs. We use scenarios of different standard deviations to simulate the changes of speeds. We find that the entropy is high when standard deviation is large and vice versa. Hence, the sudden changes of speeds can be reflected by entropy.
3. This approach is used to reflect the irregular degree of data and can’t represent the accuracy of data. That is, when the whole data of a sensor are lower or higher than the normal speed, entropy cannot reveal this fact.
4. Sometimes the value of entropy is influenced by the classification standards. In our model, these cases can be reduced through shifting the distribution of data of each sensor to the middle of the service level.
5. This approach can be extended to multi-source data fusion. The entropy is
calculated through the probabilities of classified data of the target sensor and has no relationship with the probabilities of other sensors. Hence, there is no limitation in the number of sensors in this fusion model.
6. The fusion result is close to the average of the data of sensors when the averages of each sensor are in small difference.
7. The effect of entropy is obvious when the variations of data are large. The entropy is large when the variation of data is large and vice versa. When the entropy of a sensor is much larger than others, its weight is obviously smaller than others.
5.2 The Suggestions
In this section, we offer some suggestions about data fusion for the future researches stated as follows:
1. To address weakness of entropy failing to reflect whether the row data is correct or not, we can consider another variable while calculating the weights. For example, use the historical database to acquire the average speed during a specific time period.
We can compare the average of historical data and the average of the fused data. If the difference between these two averages is large, we can adjust the entropy value.
Then the weights can reflect the accuracy of sensors.
2. We use the average of lidar data as the accurate value to compare with our fusing results. However, the lidar data is another source of traffic information and there are also errors when we acquire them. If the research time and funds are enough, the video camera may be a more accurate source for comparison purpose.
3. The classifying method can be modified. Shifting the distribution of data of each sensor to the middle of the service level can solve a portion of the problem due to the classification standards. However, there still room for modifying the
classification standards to reduce its impacts.
4. The filtering method is too simple. Since the emphasis on our model is the fusion method, we do not spend sufficient time on surveying the filtering method of the raw data. So we simply filter out the exceeding values. The fusion model will be a
more complete one if there is a better filtering algorithm.
5. The optimal weight scheme can be modified. According to some results of experiments, we conclude that the weight of a sensor is inverse proportion to its entropy. However, there may be other factors that we do not discover which influence the relationship between entropy and weight. Future researches can modify the optimal weight scheme to reflect these factors.
Reference
[1] Linn, R.J. and D.L. Hall. “A Survey of Multi-sensor Data Fusion Systems.” Proceedings ofthe SPIE - The International Society for Optical Engineering. 1-2 April 1991: Orlando, FL.SPIE, 1991. Vol. 1470: (13-29).
[2] B. Fassinut-Mombot, J.B. Choquel, ” An Entropy Method For Multisource Data Fusion”, ISIF,2000.
[3] Sarma, V.S. and S. Raju. “Multisensor Data Fusion and Decision Support for Airborne Target Identification.” IEEE Transactions on Systems, Man and Cybernetics. Sept.-Oct.
1991: (1224-30).
[4] Ministry of Transportation and Communications in Taiwan, Highway Capacity Manual, 1991.
[5] Albert C.S. Chung, Helen C. Shen, Otman B. Basir, ”A Decentralized Approach to Sensory Data Integration” ,IEEE,1997.
[6] Otman B. Basir, Helen C. Shen, “Sensory Data Integration: A Team Consensus Approach”, IEEE, 1992.
[7] Yifeng Zhou, Henry Leung, ”Minimum Entropy Approach for Multisensor Data Fusion”, IEEE,1997.
[8] Daniel J. Dailey, Patricia Harn, and Po-Jung Lin, “The Final Research Report of ITS Data Fusion”, Washington State Transportation Center and Washington State Department of Transportation, April 1996.
[9] Shu-Cherng Fang, J.R. Rajasekera, H.S.J. Tsao, “Entropy Optimization and Mathematical Programming”, International Series in Operations Research & Management Science, 1997.
[10] Ruey Long Cheu, Der-Horng Lee, and Chi Xie, “An Arterial Speed Estimation Model Fusing Data from Stationary and Mobile Sensors”, IEEE,2001.
[11] John N. Ivan, ”Data Fusion of Fixed Detector and Probe Vehicle Data for Incident Detection”, Computer-Aided Civil and Infrastructure Engineering,1998.
[12] David Keever, James Pol, ”Data Fusion To Support Information Needs for Surface Transportation Operations”, The 9th World Congress on Intelligent Transport Systems, 2002.
[13] Zhaosheng Yang, Feng Jian, Lixia Bao, ”Study on the data fusion technology in advanced public transportation system”, China’s 10th Five-Year National Key Technologies R&D Program, 2002.
[14] Daniel J. Dailey, “Sensor Data Fusion within a Regional Architecture for ITS Applications”, Proceedings of the Intelligent Vehicles '95 Symposium, 1995.
[15] Simon Nash, Ray Morris, Tom Cherrett, “Management of Journey Time Data in an Integrated Traffic Control System”, The 6th ITS World Congress, 1999.
[16] Bruce Hellinga, Rajesh Gudapati, “Estimating Link Travel Times from Different Data Sources for Use in ATMS and ATIS ”, Proceedings of the Canadian Society of Civil Engineers 3rd Transportation Specialty Conference held in London, Ontario, 2000.
[17] Shawn M. Turner, Douglas J. Holdener, “Probe Vehicle Sample Sizes for Real-Time Information: The Houston Experience”, IEEE, 1995.
[18] Bruce R. Hellinga, Liping Fu, “Reducing bias in probe-based arterial link travel time estimates”, Transportation Research Part C, 2002.
[19] Ashish Sen, Piyushimita (Vonu) Thakuriah, Xia-Quon Zhu, Alan Karr, “Variances of Link Travel Time Estimates: Implications for Optimal Routes”, Internal Transportations in
Operational Research, 1999.
[20] Hui-Wen Chang, “The Study of Using GPS Data of Buses to Estimate the Average Speed in a Link”, Thesis of Master of NCTU, 2002.