Lane-level roadway modeling is a key enabling technology in the deployment of automated vehicles and connected vehicle systems. A variety of applications for these systems would be enabled by the availability of up-to-date lane-level maps; however, current mapping solutions lack this resolution and are not dynamically updated.
Vehicles equipped with USDOT Connected Vehicle hardware broadcast positional data 10 times per second in a Basic Safety Message (BSM), designed to be used for vehicle-to-vehicle (V2V) safety applications onboard vehicles. Southwest Research Institute (SwRI) has developed processes to passively collect these BSMs through roadside equipment (RSEs) and a set of learning algorithms that utilize the information contained in the BSM to produce a local map of the roadway at an accuracy level of individual lanes. Once this map has been generated, it can be shared back to vehicles in the local area to increase their safety applications accuracy and to reduce false positive warnings. In addition, if a change occurs in the lane structures, such as a portion of one or more lanes becoming unusable due to an accident or debris, the map will automatically update once a sufficient number of vehicles have avoided the location. This method uses the behavior of vehicles themselves to determine the structure of the local roadway and enables an efficient and rapid method for keeping the map up-to-date.
Key Elements of SwRI Research
- Cooperative, decentralized approach for the generation and maintenance of lane-level roadway models (maps)
- Flexible deployment strategy for maintaining algorithms and models locally onboard a single vehicle or on a larger back-end infrastructure system
- Integration with existing USDOT Connected Vehicle devices and communications standards
- Direct applicability for use in automated vehicles and in V2V safety applications
SwRI has developed a set of software tools that enable the passive collection of vehicle BSMs to be converted into a high-fidelity, lane-level model of the local roadway structure. These algorithms utilize the behavior of vehicles, as evidenced by their driven paths, to infer the details of lane structure, which have the potential to change due to a construction lane closure, a collision, or an obstruction caused by other debris. Once the lane-level map has been reduced to a minimum set of GPS points, it is shared back to the local vehicle population, and represents the most up-to-date information on the structure of the local lanes in near real-time. The process is fully decentralized and automatic, continually updating the model(s) as vehicles drive through an area.
- Sturgeon, P., II, Avery, P., Garcia, R. A Cooperative Vehicle Application for Dynamic Lane-level Model Generation. Proceedings of the 2014 International Connected Vehicle Conference and Exposition (ICCVE), Vienna, Austria. 2014.