Understanding Information Requirements in ‘Text Only’ Pedestrian

Wayfinding Systems

William Mackaness1, Phil Bartie2, Candela Sanchez-Rodilla Espeso1 1School of GeoSciences, 2School of Natural Sciences

The University of Edinburgh Stirling University, Drummond St, Stirling FK9 4LA

Edinburgh EH8 9XP

william.mackaness@ed.ac.uk

Slide 1 of 927

Key points

• Text only system

• Rich city models required to underpin Automated Landmark Text-Only Systems (ALTOS)

• Methodologically interesting

Context

Calm, concealed technology

• …Intuitive

• …conversational

• …hands free, eyes free

𝒕𝒐 𝒆𝒙𝒑𝒍𝒐𝒓𝒆

= 𝑓(𝑤ℎ𝑒𝑟𝑒 𝑎𝑚 𝐼 +𝑤ℎ𝑒𝑟𝑒 𝑖𝑠 𝑤ℎ𝑎𝑡 +ℎ𝑜𝑤 𝑑𝑜 𝐼 𝑔𝑒𝑡 𝑡ℎ𝑒𝑟𝑒)

Research question

The ‘what’ and ‘where’ of optimal choice and announcement of landmark based instructions in Automated Landmark Text-Only Systems (ALTOS)?

Street names & Landmarks

• Street names:

– limited visibility, language constrained, absent

• Landmarks:

– Prototypical

– Considerable flexibility in use

– Function: confirmatory / at a decision point

– Preference: Social & personal dimension

Choice & Freq. of Landmark

Governed by:

• preferences, previous experiences, familiarity;

• morphological and topological complexity of the route;

• multi modal (stairs, concourses, streets);

• richness of landmarks and streetnames.

Delivery of landmark instructions

• What, when, too few/ too many…

Descriptive efficiency

The least number of instructions for the user to follow, in their simplest form

Effective Instantaneous recognition (mapping between object description and real world object)

Salient Unambiguous referent/ no confusion over what is being referred to

Robust Minimise the chance of becoming lost by including confirmatory cues

Functional requirements

1. Direct the user in multi network envt,

2. Orientate the user (so they set off in the right direction),

3. Get the user to turn/ select the right path at junctions,

4. Assure the user,

5. Let them know when they get there!

Landmark saliency requirements

mix of dynamic/non-dynamic variables:

𝑙𝑎𝑛𝑑𝑚𝑎𝑟𝑘. 𝑠𝑒𝑚𝑎𝑛𝑡𝑖𝑐. 𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑎𝑙. 𝑠𝑎𝑙𝑖𝑒𝑛𝑐𝑦 =𝑓 (𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑟. 𝑝𝑟𝑜𝑥𝑖𝑚𝑖𝑡𝑦, 𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑟. 𝑏𝑒𝑎𝑟𝑖𝑛𝑔,

𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑝𝑜𝑖𝑛𝑡. 𝑝𝑟𝑜𝑥𝑖𝑚𝑖𝑡𝑦, 𝑓𝑎𝑐𝑎𝑑𝑒. 𝑎𝑟𝑒𝑎, 𝑣𝑒𝑟𝑡𝑖𝑐𝑎𝑙𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒, 𝑝ℎ𝑜𝑡𝑜𝑤𝑜𝑟𝑡ℎ𝑦𝑛𝑒𝑠𝑠,

𝑝𝑜𝑝𝑢𝑙𝑎𝑟𝑖𝑡𝑦)

Morphological requirements

• Modelling shape of junctions (T, Y, X)

– Simple: ‘take right fork for a Y junction’, or ‘turn right at the T junction’

– Complex: exit number, turn angle: ‘take the 3rd exit heading straight on’

• Street- bend modelling, gradients, distances

Language requirements..

Adjectives left, right, sharp, straight,

Prepositions towards

Verbs Turn, walk, carry on,

Adverbs After, Before, Downhill, Uphill, immediately

Nouns metres, minutes, steps, bend, distance,

Proper nouns Streets and Landmarks

System requirements

1) 3D city model (visibility modelling),

2) Database of descriptions of landmarks,

3) Method of dynamic saliency modelling,

4) Knowledge of pedestrian location,

5) ‘Description construction’ draws upon topological and morphological form, distances, relative angles

6 km by 4.5km

LiDAR sourced DSM, DTM

DSM

DTM

Gazeteer for Scotland, OS PointX and Openstreetmap

Paths, tracks, steps n road centre lines: OSM street network

Metric Method of calculation

visible façade area Product of street frontage calculated from OS MasterMap and height from LiDAR data

viewing distance Dynamic distance of pedestrian from landmark (using smartphone GNSS)

visual unusualness Count of unique user Flickr images

function(s) Count of FourSquare venue check-ins

proximity to a decision point

Distance between landmark and ‘junction’ in multi modal path graph

prototypical form Ranked preference according to: church, monument, tower, hotel

Name recognition Recognisable wrt web search (eg ‘McDonalds’, ‘Subway’)

Street name instructions Information Landmark based instructions Go <x> metres Distance of activity Walk about <x> metres About <x> minutes Duration of activity Head <cardinal> (cardinal=west, north, etc) Toward <y> (y=streetname) Slightly right/left

Orientation Walk towards <feature>

Turn < direction > (left, right)

Network guidance terms (path descriptors)

Turn <direction> (left, right) Immediately turn <direction>

On <streetname> Locational information Stand with < feature> on your <relativelocation> (relativelocation is left, right)

Topological descriptions Opposite < feature >, next to < feature > Topographic descriptions Walk <up hill, downhill, up steep hill> Stairs, roundabout, street Object classifications Squares, public gardens, buildings, streets,

stairs Road <street name> Object descriptors Road <straight, bendy, sharp bend>

Building <stone, turrets, towers, domes> Junctions <cross roads, T junction, forks>

Object visibility < feature> is visible on your <relativelocation> (relativelocation is left, right)

Confirmatory cues You should see < feature > Continue onto <streetname> Decision point features At junction <type> turn <direction> before

<feature>

select * from wayfinding_knownstreets_jtype(2007687,2007628) select * from wayfinding_knownstreets_jtype(2007687,2007628)

Street level experiments

Evaluation of a text based system using:

• a series of geolocated landmark

Or

• street based instructions

& subject following, questionnaire, focus group

• Trajectory analysis

Information gathered • Phone:

– user input rating 1-5 of instruction

– Time taken, accelerometer, trajectory (velocities, dwell points)

• Observed:

– Lost, obstructed, search mode

• Participant questionnaire:

– Ease of interpretation

– Sufficiency of instruction

– Overall experience

– Familiarity

Experiment

• 4 routes; for each

– identify the most suitable landmarks,

– geofenced announcements via GNSS enabled device,

– announcements one at a time (avoids requirement to memorise);

Experiment

• 15 males / 15 females, no mobility impairment, smartphone users

• Via Facebook community

• Age ~ 26

• Familiarity: 2 months – 62 years

• Each subject – 2 of the 4 legs, (1.5 hrs), unaware which system they were using

Street based instructions Landmark based instructions 1. Head west on Crichton St toward Charles St - go 45 metres 2. Turn right onto Charles St About 1 min - go 94 metres 3. Turn left toward Teviot Pl About 2 mins - go 120 metres 4. Turn left onto Teviot Pl About 2 mins - go 110 metres 5. Continue onto 1/Lauriston Pl Continue to follow Lauriston Pl 6. Destination will be on the right About 6 mins - go 500 metres

* Your destination is: Edinburgh College Of Art

1. Stand with Informatics Forum on your right and Appleton Tower on your left. 2. Walk about 50 metres towards George Square. 3. Turn right before George Square at the cross roads. 4. Walk about 100 metres (with Informatics forum on your right). 5. Turn left to cross Bristo Square walking slightly uphill towards McEwan Hall (large building with a dome). 6. Turn left on to Teviot Place, McEwan Hall on your left. 7. Walk along Teviot Place continuing straight for about 100 metres. You should pass Royal Bank of Scotland on your right. 8. Carry on straight at the junction on to Lauriston Place. Walk for about 500 metres. You should pass George Heriot's School on your right. 9. After the slight bend in the road, you will go downhill. 10. Turn off right on to Lady Lawson Street, and walk for 40 metres. Your destination will be on your right, opposite the Novotel. * Your destination is: Edinburgh College Of Art

observations

Street based

landmark

Street based

landmark

Green -Street based Red - Landmark

Triangle – announcements

(purple poor, blue is good)

Circles – stoppers

(blue – few stopped, red many stopped)

Street based

Landmark

Street based Landmark

‘at the cross roads go straight towards the Black Watch monument…’

‘…Immediately turn left and descend the Playfair steps’

‘take the right fork by Novotel…..’

‘…downhill for 40m.’

Preference for the visual analog

“I am really not interested in electronic/GPS navigation – I would always just have a (paper) map. I much prefer to be more in control of where I am going and see where I am in relation with the rest of the city.”

“It’s fun to work out where you are on a map when you are exploring.”

Street vs Landmark

“Really good and easy to follow instructions. Especially useful with easy to recognise places e.g. Doctors Pub or Novotel which let you know that you are definitely going in the right direction.”

“Easy to use, very accurate and nice to have references (landmarks).”

Street vs Landmark

• Given richness of description & flexibility of delivery, landmarks preferred:

“The lack of signage in Edinburgh meant that using only street names for directions wasn’t particularly helpful.”

‘<street based instructions> Not done in a very engaging way.”

Instruction redundancy

“Instructions could have been more frequent to reassure me I had taken the right route.”

“There was quite a long gap between instructions.”

The [landmark based] systems were more reassuring because they told you which buildings you should have walked past if you were going the right way but this one didn’t therefore you could have been walking for 5 minutes and not known.”

Conclusions

• preference for landmark based instructions over street names,

• using landmarks: greater flexibility, opportunity to offer more instructions,

• some correlation -poorer ratings and complexity of route

A Broader Ambition

• The busy pedestrian dialogue interaction

• ALTOS… a way forward?

• Linking design Heuristics with database requirements. A city model that:

– needs to model the complexity of the city ‘densify’ instructions

– enables visibility to be calculated

– can rank the most striking landmarks in our descriptions

Acknowledgement The authors are most grateful for funding from the EU of the SpaceBook project.

http://www.spacebook-project.eu/