The Rail Tap app – an #ar / #ai / #cloud story
The Rail Tap app was a submission I made to the UK Defence and Security Accelerator in May of 2019.
The feedback described it as “unique”, but said that it lacked explicit support from a rail operator, and therefore clarity around the origin of the datasets that could be used, for when the Minimum Viable Product (MVP) might be tested in real time and place.
There was one other main criticism, which I shall discuss briefly at the end of this post.
I think the uniqueness of the idea is undoubted. But it’s also a mightily practical and realisable project.
It was developed with the support of two Liverpool-based software companies: CitrusSuite Ltd and Quanovo Ltd, each being experts in their respective fields.
The project itself is still very much alive and kicking: I am currently proactively looking for a rail operator and/or other interested stakeholders to resolve the challenges described by the DASA reviewers.
In the meantime, so it’s clear to anyone reading this post:
- The IP belongs to myself, even after submission to the government. Therefore, whether accepted or no in the case of future resubmission, this circumstance would remain in place.
- The opportunities to benefit from the software developed and even earn revenues – within the sector, and dependent geographies – for any stakeholder who signed up to participate in the first app trials are quite considerable:
- if delivered via a DASA contract, the app frontend itself would be free to trial;
- whether delivered via a DASA contract or not, any participating stakeholder would have the chance of acquiring the right to onward-commercialise the product/service, within the rail sector and geographies of their choice, and – depending on the agreement signed – also have the right to receive up to 100% of all revenues thus generated. This is surely too good to ignore.
To assess, then, exactly how unique this proposal was, back in May 2019, and keeping in mind the heavy conceptual development – in particular around what I have described as the intuition validation engine (that is to say, i’ve) – that has taken place since May, I have decided to partially publish the original narrative and story with which I strove to sell the idea to DASA and its client government department – ie the paying customer.
What follows below is this story.
Enjoy!
“See It. Tap It. Sorted.”
The Rail Tap app
Introduction
What the proposal will deliver
“See It. Tap It. Sorted.”, the Rail Tap app, has a simple interface, allowing users to communicate feelings about the rail network efficiently and anonymously. Specially developed AI evaluates and reports, in real time, the data which staff and public send. They receive information about the network, specific to their location, needs and roles. The security and customer experiences of public and staff are improved.Why it’s important
The app expands the “See It. Say It. Sorted.” initiative substantially, so the authorities can better predict potentially catastrophic events to the rail network such as terrorism, as well as improve customer-service delivery, with the efficient collaboration of members of the public.
The Rail Tap app story
An example of how it can work
What follows below is a simple narrative of a) how people can communicate negative feelings and emotions they have in public spaces, which quite naturally reduce their perceptions of personal safety, to deliver positive outcomes; and b) how an AI system can be designed to sort, filter and report any contradictory information received in this way.“It’s 10.30pm in July, on a station where all the staff have gone home. Mike has just got through the ticket barrier, using his train provider’s digital ticket. Alongside the standard ticketing software, his provider has given him the option of the Rail Tap module, to allow him to provide feedback throughout his train journey. He said yes, and chose to download without becoming a registered user. He now has the app on his quite standard smartphone.
Mike’s interventions using Rail Tap
Mike has arrived at the platform with twenty minutes to spare. He is now killing time. He notices that of the five lights on the platform, three do not work. He sends a notification using the Rail Tap module: “Three lights still not working at Prescot station!” He then takes a photo of the offending lights. Rail Tap has a slider button which moves from green happy to sad red. He submits the photo firmly on sad red. The photo automatically contains location data. (However, it’s worth noting that once submitted, all content can be automatically removed from users’ devices.)
Suddenly, in the far corner of the station, where the lighting is poorest, he notices a movement. He sees a flash of something. He begins to get worried, as he is the only person on the platform. He takes another photo of the wider station, as if just snapping for pleasure. He adds a social network-style “sticker” (a computer-generated object which you can paste on a photo, and which is a widely used communication strategy) to the far corner of the photo, near where he saw the flash, before then submitting it: the sticker is a picture of a cartoon dagger.
The AI’s response
The AI, a deep neural network on an on-premise server (depending on the wishes of the client, public cloud could also be used, as could any hybrid) of Northern Railways, receives in real time the three notifications from Mike. The first one, the text message, is classified as a maintenance issue, and is pushed to Maintenance to be dealt with. The second one, the photo of the broken lights, is used by the AI for two purposes: a) to support the validity of the first notification, in order that the likelihood of a false flag be discarded; b) so that the AI’s deep neural network can continue to learn better to identify what a broken light specifically on Prescot station looks like.Officer Brett and the BTP officers
On receiving the third notification, the AI raises an alert. A Northern Railway train is minutes away, and two British Transport Police are on it. The AI immediately notifies the Control Hub, where Officer Brett is on duty, of a potential threat to life at Prescot station. Officer Brett evaluates both the photo and the information. She notifies the BTP officers that someone may have an offensive weapon.
Gina’s interventions using Rail Tap
The AI receives a notification from a second user. Passenger Gina has just arrived at the station. She sees a young man in a corner of the station, using a knife to lever open the drinks’ machine. “Homeless person, dehydrated on Prescot station. Has a knife. Breaking the drinks’ machine,” she texts. The AI receives this notification, compares location and content with Mike’s earlier photo and sticker, and offers Officer Brett an evaluation based on previous events at similar stations. She decides to push the raw data to the BTP officers: Mike’s second photo with the sticker, and Gina’s comment. (Both pieces of content are anonymous, as neither had chosen to be a registered user.)
With all the above information prior to arrival at Prescot, the officers are forewarned about what to expect, and better prepared to deal with the young man.The Rail Tap and AI system’s impact on security and customer-service delivery
Mike, meanwhile, already felt better for sending the notification, as his experience in the past six months shows that changes are made in response to the data. He hopes to see the broken lights mended shortly, especially in relation to the unease he reported this evening.
Gina feels good about telling the authorities that a person at risk needed help.
Additionally, Officer Brett has been supported by the AI into delivering a more secure and customer-focused service: at no stage in recent implementation has she ever felt she was going to be substituted by the technology. Instead, she feels empowered to do her job more competently, in an increasingly cost-driven environment.
Finally, and most interestingly perhaps, the AI compares the information received in the time described above, and notes that images sent by passengers feeling considerable unease have been received from multiple stations recently, at the same time as notifications of poor lighting late in the evening. It decides that some maintenance issues need to be labelled as impacting on security, and sends a second notification with a security label to the Maintenance Hub for this to be evaluated by domain experts.”
Before I end today’s post, here’s one final interesting – at least for me – observation. One of the most negative pieces of feedback the submission received was that I hadn’t demonstrated an advancement in science or knowledge.
I feel, personally, truly, sincerely, that the above narrative, practically the first thing the reviewer would have read, shows – whilst with an arts-based mindset I accept, a mindset that may not have coincided with what the reviewer was most comfortable with, and yet with intelligence and clarity nevertheless (whatever one’s prejudices around the thinking processes employed) – that advancement will take place with this and other i’ve projects, and could in fact already have been happening, if the project had received the due go-ahead in July of this year.
i've 
You must be logged in to post a comment.