On Engineering Embedded Security Systems for Our Future…
The AI is not replacing Processors and Reviewers, it essentially makes their jobs easier. When I first came along a Reviewer could really only look at about 40 to 50 events per day. Through a lot of modifications to architecture and process refinement, the efficiencies were increased and we were able to have a Reviewer look at 1400 to 1500 events per day per person. "
As VP Software Engineering focused on Embedded Security Systems, Matt’s been involved in every part of the hardware to software process.
For the better part of the last 50 years school buses, and the technology that moves those same buses, have remained the same. It’s almost as if the entire world has progressed but the transportation for our future generations has lagged behind.
This reality is certainly not indicative of the urgent concern coming from communities and law enforcement across America. On the contrary, school boards, bus operators and transportation departments work tirelessly to serve their communities and stay on top of ever-growing safety concerns. Police work to protect our communities and keep up with the near exponential increase of automobiles speeding through thoroughfares, school zones and undoubtedly by stopped buses.
Adapting to a changing world.
Although technology has improved learning in the classroom and extended learning to our homes via smartphones and computers, school buses remain notably out of step with progress. This is in great part due to the lack of funding for schools.
Society is now at a point where the logistical challenges of fleet management for students can be addressed with modern technology; however, cost barriers remain and prevent communities from modernizing and upgrading their severely outdated fleets. The result is unacceptable stagnation — our school buses still have 1960’s technology.
Software Engineering Embedded Systems is a puzzle with evolving pieces
The pieces of the puzzle are scattered and the question remains: how do we implement this on a national scale to address the 17 million illegal school bus passing? BusPatrol developed both a model and solution. The model is simple: get the drivers who illegally pass school buses to fund the programs to keep kids safe. The solution is far more complex and requires deep innovation, AI, LTE, GPS, multi-angle camera systems and advanced round planning. More than that, it requires a dedicated and ever-expanding team of problem solvers with the creative ability to leverage systems and create never-before-seen code and tech.
Introducing Matt Farmer: Innovator. Communicator. Leader.
Leading the way on this safety tech development and the implementation of new and innovative systems to manage the problem is Matt Farmer, VP of Software Engineering at BusPatrol. A 10 year veteran in the student safety space, Farmer and his team are taking 17 million problems and crafting 1 elegant solution to perfection; an IoT ecosystem centred around a unique application called AlertBus™. Before working to create AlertBus, Matt attended the New Orleans at the University of New Orleans. As an undergrad, he majored in Computer Science and specialized in Embedded Systems in Security. He then completed his Master’s Degree in Information Assurance and Security.
I had a chance to speak with Matt to get a better understanding of how BusPatrol operates on the cutting edge of innovation. Now a well-funded startup, Matt is expanding his Software Engineering Department and pushing the boundaries of what is possible for student safety.
Understanding the problem and solution via the Event Pipeline
JOHN: So Matt let’s start talking about the problem itself. Vehicles pass school buses at an astronomical rate — it’s mind-boggling. No one had developed the technology to tackle this problem until you and your team came along. You started building AlertBus from the ground up. Tell me about the whole process. Can you explain how the process and how the technology works?
MATT: It’s a large feature set, but I’ll start with what we call the event pipeline. When a school bus stops and extends its stop-arm, that’s when we start generating what we call an event or a potential illegal pass. This is when kids are getting onto the bus or offloading at home or at school, you know, both places. When the school bus stops and the arm extends. That actuates, the relays in the bus’ computer. The information including the time, date, GPS location. The video for that event gets marked and then transmitted via a 4G LTE cellular connection over a private connection to our cloud. With this, the data is ingested into AlertBus, that’s when the classification process starts.
Understanding the Ecosystem
JOHN: So this system you’ve taken part in building is the bridge between hardware and software. As VP Software Engineering you’ve been focused on Embedded Security Systems. This means you’ve involved in every part of the hardware to software process. You essentially handle and guarantee the flow and organization of data.
MATT: Yeah, It’s definitely a team effort. At one point or another, I’ve been involved in every part of the hardware to software process. The hardware, cameras, GPS and onboard computer capture data, then the software says, ‘hey, first off, there’s an event, where or when was this event occurring? How many kids were there getting on the bus?’ And then there are also geo-fenced locations around different areas that are both ticketable and non-ticketable locations.
Why video is important in motorist education and the Event Pipeline
JOHN: So the data is not only time and location-based but, of course, includes video?
MATT: Yes. The videos from the cameras on the outside of the bus then get input into what we call AVA, our AI software, that’s how we identify events. She’s kind of our machine learning and computer vision pipeline. She does a lot, but for just this part of the pipeline, what she does first is go through the videos that get transmitted for each event. Then she does some recognition answering basic questions.
JOHN: So like, are there cars there? Are there cars passing the side of the bus, and which direction are they going in?
MATT: Yes, so AVA then takes that video, extrapolates that information, and then makes a decision on whether or not there were cars even present before a human starts looking at the video. All that information is then sent into the AlertBus system. It’s then classified into different areas in the interface for it to send information back to the school bus. It basically says, “something may have happened here, there is an event, send us the rest of the videos.” So then the buses transfer the rest of the data. There are seven different cameras that it can request, and send back to AlertBus.
So the videos come back and then what happens next?
MATT: It then gets put into the AlertBus pipeline where it moves on to the Processor stage.
JOHN: So this is where the Processor, a human Safety Expert, looks at the footage and data?
MATT: Yes, the Processors then go through the combination of all of those videos and build what we call an evidence package for law enforcement. Only the police can determine if a violation is made. They can leverage a secure platform with an easy-to-use UX interface.
AlertBus: The UI focal point for our Ecosystem’s Embedded Security Systems Event Pipeline
JOHN: So it seems you really have different systems working together and all funnelling into AlertBus to facilitate ease of use for the officers. The amount of data must be enormous — what are some of the challenges dealing with all of this?
MATT: Well, there’s been the… you could call them the ‘technology challenges’ first and foremost. I can remember a long time ago we tried to start this whole thing with a cell provider. At the time it was on what was a 2G Edge Network. You can try and imagine all of the data that’s flowing back and forth. We evolved from 2G edge, up through 3G, and now we’re 4G LTE, and we’re looking forward to 5G and wideband. Now we are able to manage more data and process things faster in higher video resolution. JOHN
BusPatrol’s seamless integration of hardware and software engineering for embedded systems.
JOHN: So in time, the hardware and communications technology caught up. The software solutions you were developing and dreamed up, for engineering embedded security systems could become a reality.
MATT: That’s right. The technology progression has really spurred a lot of innovation for Reviewers and Processors. We’ve had to build the architecture of our systems to facilitate AVA. Over the last few years, we’ve incorporated data annotation so AlertBus users labelling video can teach the neural networks to learn from those human decisions. In the past, automated enforcement companies have leaned heavily on human intervention. We’ve been fine-tuning the AI over the years.
We’ve taught it and built it to try and work out exactly what law enforcement wants to see to make their determinations. The AI is not replacing Processors and Reviewers, it essentially makes their jobs easier. When I first came along a Reviewer could really only look at about 40 to 50 events per day. Through a lot of modifications, architecture and process refinement the efficiencies were increased. Now we’re able to have a Reviewer look at 1400 to 1500 events per day per person.
Our Future is in Embedded Security Systems
JOHN: To get AlertBus and the supporting embedded systems to where they are must have been quite a journey. So how did you get there? You must have a team of coders and developers working in different languages to manage and run the flow of information on the Software Engineering side.
MATT: As far as languages that we’re working with, it’s kind of spread across different parts of the platform, you know, whether you’re working in one of the online applications, or if you’re in the backend. So for the different languages that we’re using, I guess, we’ll start mainly with Ruby and Ruby on Rails. Those are the core applications that we’ve built over the years, namely, AlertBus and Console. And there we also utilize Vue, it’s a JavaScript framework on the front end. As we’ve been upgrading the platform over the years we’ve moved to Ruby on Rails. We’ve continued with different types of HTML generation to using JavaScript on the front end. That’s different from way back in the day when we used to be a PHP shop.
If you move over to Ops, we’re using a lot of Python on the backend for scripting, Cloud Watch and Cloud Formation. And then if we move over to the AI side, we’re also using a lot of Python. So really, just depending on where you’re getting your hands dirty in the applications, just which language you’re using there.
Thank you for your time, I know you don't have a lot of it to spare!
JOHN: Matt, I want to thank you for taking the time out of your day to talk with me. We covered so much, I think we’ll pick this up again next month. I’m looking forward to following up on Software Engineering Our Future: Embedded Security Systems. I’d love to know more about how you started in tech and how your team is growing.
MATT: I look forward to it also. We’re quickly expanding our team here as we bring more and more buses online across the country. It would be great to speak about what we do on a day-to-day basis at BusPatrol.
