TrafficCast
For more than a dozen years, TrafficCast engineers have been researching the science behind driving, how traffic flow works and how congestion develops.
Data derived from intelligent transportation systems, advanced traveler information technologies, global positioning systems, cellular probes, weather radars as well as good old fashioned cameras and on the road tipsters inform complex, proprietary algorithms and synthesis models to predict traffic speed and flow.
TrafficCast predictive flow modeling employs a combination of statistic (including Bayesian), heuristic, simulation and Dynamic Traffic Assignment (DTA) analysis to address various scenarios based on availability of different data types.
- Statistical models predict traffic flow characteristics by identifying regularities in traffic flows and traffic flow patterns over time, estimating future values from the immediate past.
- Heuristic models adapt or learn optimal applications of statistical analysis of particular trends and circumstances. Fuzzy logic and neural networks are classified as part of heuristic models.
- Simulation models use random numbers to predict various factors such as desired speed, red/yellow traffic light signals, and gap acceptance (acceptable time interval between oncoming vehicles to enable an "opposing" vehicle to cross the roadway).
- DTA models are used to coordinate the optimal route assignment among motorists while maintaining the feedback of real-time traffic congestion to travel time changes and traffic information provision.
Information Input - Breadth and Depth
The foundation of TrafficCast IP is that
- It is critical to collect useful data, not just any data available publicly or commercially
- Experience enables accurate normalization of data from different sources
- A traffic engineering perspective affords the best use of collected data, both for modeling input and also as a benchmark to calibrate analysis in near real-time (an example is described in the later section).
TrafficCast is at the forefront of traffic information technology, researching and incorporating floating vehicle data (FVD) from GPS devices, Bluetooth, Wi-Fi Hot Spots, and cellular probe technology. Adopting probe data research from the ADVANCE Project (the largest traffic probe demonstration project in the US from 1991-1996), TrafficCast research has contributed to our affiliate company, TrafficCast China. In partnership with China Mobile, they have successfully deployed a travel time information system in real-time using cellular probes in Shanghai, serving more than 14 million China Mobile subscribers.
TrafficCast uses a rigorous Quality Assurance and Quality Control (QA/QC) mechanism to calibrate and verify data. Proprietary TrafficTracker™ measures data beyond metrics to what drivers see through the window. And model output is constantly refreshed by new information from FVD and public sensors, both for real-time verification as well as analytic adjustments.
Consideration of Driver Behavior
Uniquely, TrafficCast also considers driver behavior. For example, rain affects drivers in Chicago quite differently than it does in Los Angeles. Driver types and vehicle classes are also important variables. Is it fair to say that drivers in New York are more aggressive than those in Tampa? That truck drivers have different travel patterns than commuters? TrafficCast takes these factors into account.
TrafficCast expertise and knowledge in modeling and transportation engineering reflect many years of experience in research and real world traffic engineering projects. TrafficCast knows how data should be collected, fused, produced, and used. TrafficCast co-founder, Dr. Bin Ran, PhD, wrote two of the standard graduate student texts used in the transportation engineering departments in universities around the world. His work has set the foundation for modeling dynamic traffic network behavior around the world.
TrafficCast holds a number of patents in traffic information gathering, processing and distribution, and is in the process of applying for additional patents to protect its data collection, processing and fusion systems.