This document has some evaluation criteria and open questions regarding the data schema that ATE is designing, as of spring 2023. ATIP has already been used to collect data about planned active travel schemes for ATF4, but aside from mapping the geometry of interventions, we imposed only the simplest categorization -- is it a route, area, crossing, or something else? Going forward, we want to collect (and analyze) much more structured data -- but what specifically?
How do we know if a schema will do its job well?
- Can ATE inspections and data science teams use the data to do their job?
- Is it sufficient? If it has too much detail, is it easy to downsample and present a simplified view?
- Is it easy for LAs to fill out?
- This will partly depend on the UX of ATIP
- How well do the ATF4 submissions fit into it?
- Note that we don't need to capture absolutely everything -- a free-form "other" category is always necessary to keep
- Can the schema gracefully evolve over time?
- Generally, collecting more details about a type of intervention or collecting new types of interventions entirely is easy
- Changing the details collected about an intervention may be harder
- Changing how we reference existing road geometry in the presence of dual carriageways may be very difficult, for instance
- How much does the schema bidirectionally map to and from existing data, such as OpenStreetMap?
Ultimately, one of the best exercises to test a schema is to map real schemes with it.
Largely deferring these for now; we need to figure out what data to collect, rather than how to store and manipulate it.
Nonetheless, for everything we might collect, we could imagine ATIP guiding people through filling out details for it. Imagine different route segregation or crossing types. For each case, we could ask:
- How should we symbolize / visualize it on the map in ATIP?
- How do we describe it?
- When does ATE recommend this choice? What references to LTN 120 and other design standards can we point to?
- Approximately how much does this cost?
- Of course this could depend on many factors we don't want to focus on for data collection, but can we give rough estimates?
- Do we have some real-life example photos to show?
- "Part separated" vs "stepped" vs "advisory lane" route types could be much easier to see than to describe!
Many of the schema drafts are expressed as a JSON schema right now. This is an implementation detail at this stage; it's not yet the time to decide how we'll serialize data in a database, as a GeoJSON or similar file export, etc. It'd be fine at this stage to write a free-form text document saying something like "width must be a positive decimal value in meters." If somebody on inspections wants to mock up a schema, something user-friendly like Google Forms would be a fine option.
We're by no means the first group to go through this exercise of creating a taxonomy for active travel infrastructure. We should learn from, and possibly be compatible with, other schemas like:
- OpenStreetMap (OSM)
- For example, https://wiki.openstreetmap.org/wiki/Proposed_features/separation is a newer page about types of physical segregation
- This covers all types of infrastructure and has worldwide scope; we have a more limited domain
- osm2streets
- This is an attempt to represent street networks with sufficient detail to run a traffic simulation: individual lanes, turning movements, traffic signal timing, etc.
- Its purpose is also to simplify and normalize OSM data.
- TfL's Cycling Infrastructure Database
- Ordnance Survey
- ATF4 guidance
- See https://www.gov.uk/government/publications/how-to-complete-the-active-travel-fund-4-proforma/guidance-note-for-local-authorities-to-support-completion-of-the-active-travel-fund-4-proforma#types-of-infrastructure
- The questions asked are usually just "is this high, medium, or low complexity?" but this is computed from other details.
These're some of the bigger questions I want to resolve.
- What do we need to know about nearby streets?
- Do we need to know about every side road that intersects a proposed route?
- What about roads one or two intersections away? Do we need context on everything near an intervention?
- What do we need to know -- width, speed, lane configuration (including turn lanes)?
- How do you reason about dual carriageways and split roads?
- Blackfriars bridge
- St George
- Borough station
- Should our schema include contextual data that changes over time?
- 85%ile speed and average traffic count
- Does our schema include both details we ask LAs to input and details that we
can automatically calculate?
- I'd say yes; over time, we can automate "existing width"
- For 2023 (v1 or so timeline), how much detail needs to be in ATIP and part of
automated analysis?
- When would you expect/prefer to open a CAD or PDF file manually to answer something?
- Do you ever need to look at traffic signal timing plans (from the main junction or nearby)? What do you use them for?
Many details here are also in Github issues or part of a JSON schema draft.
- Types
- on the road - parallel to a road
- dedicated off-road, ie through a park
- canal towpath
- Users
- Cycle or foot only?
- If shared, any marked separation?
- Cycle lanes on road
- width (min, max, average, exact geometry)
- surface type
- segregation on left and right (from moving vehicles, parked cars, buses, pedestrians)
- Upgrades to existing routes
- resurface (fix muddy path)
- widening -- from what to what?
- add CCTV and lighting along path
- grade separated
- bridge
- tunnel - at-grade
- signalised
- pegasus
- puffin
- toucan
- need to press beg button? or auto-activated by sensor? or on a timer?
- max time between pressing button and getting green?
- how long does green last?
- priority
- zebra
- school?
- parallel?
- uncontrolled
- refuges
- dropped kerbs
- tactile pavement
- does it match pedestrian's desire line?
...
- advanced stop lines / bike box
- how many seconds between the bike and car signal going green?
- modal filters
- point or diagonal?
- physical implementation: planter, bollard, raised kerb, bus gate / camera
- exceptions to entry? residents?
- parking (racks vs hangar, capacity)
- pavement extension / bulb-out
- side road treatment / raised junction
- wayfinding signs
- markings on road
- new parking restrictions