Institute of Transportation, MOTC

       In continuation to the research results of the preliminary, in addition to continuing to improve the data interfacing of data sources from respective units and strengthening the “traffic event integration and information circulation service platform”, this plan also continues to improve event management platform related functions and enhance the limits of authority for operational security. Additionally, in order to enhance automatic traffic event detection capacity, this plan carries on previously installed AI detection devices. Through the formulation of regional planning logic and situational tests on detection angles and shooting distances, the best mode of recognition can be grasped. On top of it, different detection items are defined based on defined road junctions and sections to obtain accurate traffic flow parameter detection rates (in terms of traffic flow accuracy, motorbikes and passenger cars maintain a rate of above 90%; large vehicles reach above 80%; vehicle speed accuracy reaches above 80%; lane occupancy rates reach above 75%). Furthermore, through AI learning framework and detection logic adjustment, AI detection devices are used to detect multiple event s(including accidents, unusual pullovers, congestion, and flooding events), enhance event detection accuracy rates (unusual pullover detection rates ranging from 80%~98%) in order to quickly obtain information and images of event occurrences. In addition, the standardized API will automatically detect events and transmit them to the event management platform. In order to present related implementation effectiveness, this plan conducted traffic event impact analysis using probe vehicles and data of AI image recognition of vehicle speeds. The correlations among event occurrences, event exclusions, and vehicle speed changes were grasped to evaluate the impacts of events on traffic flow. Furthermore, the accident cases were also used as examples in this plan to estimate and evaluate improvement effectiveness before and after platform importation. Results show that after the importation of the event management platform, the time for handling a single event was reduced by about 14 minutes. It is estimated that the saved benefits after one year of travel time amounted to NTD1.8 billion; the benefit of gasoline consumption reduction amounted to about NTD3 million; the benefit of carbon dioxide emission reduction amounted to NT$7 million. At the end of the term in the plan, the plan implementation effectiveness was advocated through publication of relevant articles, conduction of result and experience sharing briefings, and video production.