Open cody-somerville opened 12 months ago
Thanks so much for the detailed write up. This is super helpful. It seems from the linked pdf that the number of such connectors is small. Do you know if there are more situations across the state/country? Thanks
Hi Dennis, thanks for the quick reply.
The document linked was from TXDot's Austin Office so is likely scoped to that area. I'm not a subject matter expert but I wouldn't necessarily say the number is small. For example, these lanes typically exist at cloverleaf exchanges which was a previously popular design in the United States. See TXDot Design Manual and MUTCD.
Based on a Texas A&M Transportation Policy Research Center publication, target market for this type of investment is larger freeway interchanges where lane changes cause mainlane traffic to slow down unnecessarily; and freeways in areas with closely spaced major roadways and little room for entrance and exit ramps.
Also, like with most things traffic, there are many variations. I've also found examples where the bypass exists on the outside instead of the inside like here in San Antonio: The mappers in this case have both ways tagged as highway=secondary and we see it routes via the bypass though it is unclear if that is a fluke or not.
In Ontario, Canada, sections of the 401 Trans-Canada Highway around Toronto have expressway and collectorways. Both are mapped as motorways as they fit the definition. Both have the same maxspeed. It seems however that OSRM prefers to stick to the collectorways (though will occasionally crisscross). Local knowledge suggests that which is faster is usually the one you aren't on (lol) so I can't outright recommend that routing blindly via the "express" is superior but it is the design intent and was my initial expectation. Joking aside, except during rush hours, traffic on the express is generally 10-20km/h faster than collector (which yes may be over the speed limit).
Issue
The routing engine may currently overlook optimized paths through Collector-Distributor (C-D) roads, leading to sub-optimal route suggestions especially in scenarios involving signalized intersections which these C-D roads are designed to bypass. These bypass roads, as per TXDOT's definition, are extensions or splits from continuous frontage roads providing a faster alternative to reaching the motorway mainline or bypassing signalized intersections.
In many instances, these C-D roads are mapped as secondary_link and often lack a tagged maxspeed, though their speed limit is inherently at par with or higher than the frontage road they branch from. The absence of this data leads to a preference for frontage roads over the C-D roads in the current routing logic, thus potentially suggesting slower routes to users.
Additionally, in scenarios where a motorway on-ramp is only accessible via the C-D road, OSRM may skip the entrance to the motorway entirely, routing instead to the next accessible entrance that isn't via a C-D road as typically mapped. This overlooks the intended utility of C-D roads in providing expedited access to motorways and could lead to significantly longer and less efficient routes.
Example in Austin, TX:
In this second screenshot, the start point is moved slightly onto the C-D road to force it to route via it and results in optimal routing for demonstration purpose (you'll notice alternate routing that returns to frontage road displayed as well)
Note that I updated examples above to set maxspeed so they'll soon route differently then visualized. I'll follow-up with example that I'll leave in typical tag configuration to act as example.
In some cases the C-D road is mapped as secondary and the frontage road with intersection is connected with secondary_link such as this example also along I-35 in Austin:
Feature Request
This enhancement request aims to capture and address the routing pattern associated with Collector-Distributor roads, aligning OSRM-backend's routing logic with real-world road network dynamics, and thereby offering more optimized route suggestions to users.
The routing engine should preferentially select C-D roads over frontage roads when the former provide a faster route option, especially around signalized intersections which they are designed to bypass, or when they are the only means to access motorway on-ramps.
Heuristic Enhancement: Develop a heuristic to identify C-D roads based on their common characteristics and the local road network structure, even in the absence of explicit maxspeed tagging.
Maxspeed Inference: In cases where maxspeed is not tagged, infer the speed limit from the adjoining frontage road or use regional defaults for secondary_link types as a baseline.
Routing Logic Adjustment: Adjust the routing logic to favor C-D roads over frontage roads, especially when the former helps in avoiding signalized intersections, or providing the only access to motorway on-ramps, thereby improving the overall trip time.
Additional Resources
TXDOT Collector-Distributor Fact Sheet