5. Conclusions and Outlook
In our survey, we have studied different traffic flow prediction models in depth, motivated by their possible contribution to ATMS and ITS systems to forecast potential traffic conditions, thereby solving traffic 645 management problems in smart cities.
In Section 2, we examined currently available data sources used for traffic flow prediction. In our opinion, the enumerated data sources should be used together, because every data source has its own advantage. That way, one can achieve the best result by fusing them in an appropriate model.
Among fixed position sensors, the sensors able to scan more lanes at the same time (e.g., video image processors or laser radar sensors) could be more cost effective than other fixed position solutions. These can be used to implement crowd surveillance tasks in cities as well, so they could be the eyes of future smart cities due to their versatility and flexibility.
With moving sensors, we can identify exact paths, speeds, and moving patterns of vehicles and pedestrians, which can reveal direct connections between adjacent road segments. Moving sensors have minimal infrastructure cost compared to fixed position sensors, and they are important data sources in areas that are not covered by fixed position sensors.
