Loading...
Please wait, while we are loading the content...
Real-time Urban Road Network Travel Time Prediction Based on Probe Data: Framework and Stockholm Case Study
| Content Provider | Semantic Scholar |
|---|---|
| Author | Jenelius, Burghout |
| Copyright Year | 2016 |
| Abstract | 1 The paper presents the architecture and methodology for an urban road network travel time predic2 tion framework based on low frequency probe vehicle data. The prediction framework is intended 3 to be used in a real-time traffic management and information provision tool. The design of the 4 framework has to satisfy three crucial aspects: computational efficiency of real-time prediction; 5 flexibility to changes in the network; and robustness against noisy and missing data. Correlation 6 patterns between links and time intervals are computed, updated and stored at infrequent inter7 vals based on historical data and a Probabilistic Principal Component Analysis (PPCA) model. 8 Prediction for future time intervals is performed in real-time using stored model parameters of 9 correlation patterns and recent current-day observations. A novel multivariate hybrid method of 10 PPCA and local smoothing, which considers local correlations among neighboring links, is pro11 posed. The methodology is applied to the road network of Stockholm, Sweden and probe data from 12 taxis. Computational experiments show that the proposed method provides high accuracy for both 13 the peak hours as well as off-peak traffic conditions. The effects of missing data and link speed 14 variation on the prediction accuracy are also analyzed. For links with large link speed variations 15 in particular, the PPCA based methods significantly outperform the historical mean, whereas for 16 links with high proportions of missing data the hybrid PPCA method is more reliable than pure 17 PPCA due to its ability to use local neighboring link correlations. 18 19 |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://people.kth.se/~jenelius/CJB_2016.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
