Transport for NSW has launched a new ‘skill’ for Amazon’s Alexa voice assistant that offers a swag of new features for commuters, as the agency continues its push towards intelligent automation.
Head of technology and innovation Chris Bennetts told iTnews the new custom-developed Alexa skill was rolled out in late June, partially in response to the coronavirus pandemic.
The bot is an evolution of the previous Alexa skill, dubbed Transport Bot, which was released in early 2018 based on the agency’s Real-time Intelligence Transport Assistant (RITA).
“What we’ve done recently is update that skill in Alexa and enable a whole treasure trove of information to be able to be spoken to,” he said.
Bennetts said this included patronage data, which is collected from the Opal card system on buses and a mixture of Opal data and weight data sourced from sensors on trains.
That real-time data, which has been altered to reflect physical distancing in recent months, is already fed to third-party apps such as TripView, AnyTrip and NextThere.
Unlike the 2018 Transport Bot, commuters can use the new Alexa skill to plan a trip or journey by surfacing the next available public transport service from any given location.
Bennetts said commuters could simply ask, for example, “hey I want to know when the next service is leaving from Central to North Sydney”.
“And not only will Alexa tell you but then you can ask it how full that service is from a social distancing standpoint,” he said.
A quick test of the skill shows that it relays and displays occupancy as low, medium or high. Information about fares, wheelchair access and trip alerts is also available through the skill.
On third-party apps this information is displayed using either a ‘jelly baby’ or traffic light system depending on the mode of public transport used.
The skill can also be used to plan trips beyond train stations, bus stops and ferry wharves.
“If that service has been disrupted, you can understand if there's a rideshare operator available as part of your trip or journey as well.”
Bennetts said that while the agency had always planned to bring the new features to Alexa, the pandemic accelerated this process for what remains an “emerging channel” for TfNSW.
“We were always going to do it because we want to bring parity of experience across different channels, and it's just about lining things up and getting it done,” he said.
“And by making all this social distancing and occupancy information available through open data we were able to very quickly bring parity of experience to that channel.
“The great thing about our strategy around opening up the data, opening up our APIs and making a consistent experience, whether it be an owned channel or third-party channel, is that when that disruption happens customers will be able to consume and use transport information.”
AWS the centre of real-time, predictive data
While TfNSW started off using AWS as part of its shift to IaaS six years ago, it has since expanded to consume 87 AWS services to enable a better travel experience for customers.
“We host our open data on AWS, so that’s literally stored in an S3 bucket,” Bennetts said.
“All of the real-time data across the network [is] available through an API gateway where developers can pull that and publish it into their own applications.
“We’ve also got our traditional journey planning capability through our transportnsw.info asset and the Opal Travel app all hosted on AWS.”
The AWS stack is also central to TfNSW's ability to understand and predict occupancy across the public transport network.
This is done using both using the Opal card system and in-built sensors under each axle on a Waratah train carriage. Where real-time data is not available, historical data is used.
The data is stored in S3, and machine learning (ML) algorithms are used to make predictions.
While TfNSW at one point used the ML managed services platform Amazon SageMaker, it has since moved to deploying and running ML models with AWS Lambda.
“Previously we were using SageMaker just to see what was possible from a machine learning standpoint around predictions … based on weather … or any kind of special event data," he said.
“But I think we’re able to just do it with just Lambda at the moment, which is quite exciting.”
Bennetts said that TfNSW is mostly using AWS and Microsoft Azure for public cloud services, but intends to take a pragmatic approach to the use of all major cloud providers in the future.
“What we’ve found, traditionally, over the last few years is that different platforms have different developers available for them at different times,” he said.
“So in the last four or five years it has been pretty easy to pick up the phone and get some AWS developers in - it’s been harder to do that on other platforms, but that’s equalising a bit.
“Our future kind of looks at using the best of all of them and converging them in a smart way, so that we’re really seeing what’s possible with each of the applications.”
Intelligent automation ahead
Bennetts said that, looking ahead, TfNSW planned to introduce more real-time sensors across the network to build on recent developments like its operational dashboard built by Anytrip.
But he wouldn’t be drawn on whether this included retrofitting older fleets of trains like the Millenium fleet.
“What we’ve learnt through the COVID crisis is how the power of data can be used to drive operational decision-making,” he said.
“So what we’re really interested in doing now is doing more with real-time sensors across the network, so we can get more and more data out of the system.
“And then use that and build intelligent decision engines that give our operational teams greate next best actions on what to do across different scenarios.
“No one knows the system better than the operators, but what we want to do is to use intelligent decision engines to help inform them and give them better choices.
“So that’s super interesting to us - how we can use these technology stacks to do that.”