Skip to main content Skip to navigation

FAB wayfinding system

Problem: people struggle with finding locations in the building, there is very little signage (or scope for adding signs), and in some ways it is a confusing building with little uniformity between floors.

Solutions: we are create a FAB App that will provide directions to a chosen location.

Challenge: what is the best way to give directions for the FAB? how can we help people to follow the directions they are given, being able to check that they are following the correct route as they go along?

Solutions: the floors of the FAB do not have clear enough zoning (for example, study space 1 on floor 3) - design a zoning system so that we can say to people "go to zone x on floor y and find your location by doing z; there are not enough well known named landmarks that can be used in directions, develop some; use photos, video; develop an approach that the app can use to know where the user currently is in the FAB - we have looked at image recognition, QR codes. Zoning and landmarks will help with this; do we also need to provide a map or 3D model?

Experiment 1

First I wrote a detailed description of the first three floors of the FAB, and then got Microsoft Copilot to write instructions, as a series of steps, to guide someone from one point in the building to another. The results were good, but they pinpointed mistakes in the description, misleading, and missing information. I then edited the description, repeated, until the AI generated instructions were good enough. I could then generate instructions, and use them to build a web application. I used one moderately complication route as a test. With the help of GutHub Copilot (the coding tool) I quickly created and tested a prototype. The instructions were turned into an MP3 using Eleven Labs Text to Speech. And used to construct the prototype. Voice control tells it to move on to the next step, which is then read out.

The prototype is live online for testing. It might get turned into a second version that can take sets of steps as a json file, for many different routes, and the .mp3 auto generated using Eleven Labs API. The AI generation part of this would only happen once per set of directions.

Try it out (turn your sound up, and grant mic permissions when asked): https://inspireslearning.com/navigation/Link opens in a new window 

Feedback, bright ideas, offers of help to Robert O'Toole please.

Let us know you agree to cookies