Georgetown Law
Technology Review

In an interview with Georgetown Law Technology Review, Dharmesh Tarapore, founder and CEO of Aircraft Collision Avoidance System (ACAS) Technologies, discusses his tech start-up journey and the technology behind his flight collision avoidance system.

Tarapore’s invention utilizes machine learning to increase flight safety for general aviation airplanes by detecting and alerting pilots to incoming aircrafts. The technology leverages highly technical mathematical computations to predict an incoming plane’s trajectory and then subsequently suggests to the pilot an alternative route to avoid a mid-air collision.

Tarapore, who also teaches computer science practicums part-time at Boston University, plans to have the first thousand units of his technology shipped by the end of this year.

GLTR: Hi Dharmesh, thank you for taking the time to discuss your company today. I would first like to ask you, what motivated you to start ACAS Technologies?

Tarapore: I got my private pilot’s license in 2016 and I flew everywhere that I possibly could. I took this tiny little airplane and flew to places like Vermont, Canada, and on one particularly memorable occasion, Florida. I was flying back [ to Bedford, MA]  from Portland, Maine and air traffic control warned me about an airplane in my vicinity that was not talking to anyone. Air traffic control said they could see this plane approaching mine on their radar and to be very careful. I was flying with a friend, and we started looking frantically outside the window and we could hear air traffic control’s proximity warnings going off in the background on the radio. So, we were looking for it, but we did not see it until it was too late. The aircraft missed us by a little less than 200 feet and it was sheer dumb luck that we escaped unscathed. The other airplane was so close that I could almost make out the writing on the pilot’s t-shirt.

At that point I didn’t know if I was ever going to go flying again, because what happened scared the hell out of me. So, I started doing some research and learned about TCAS 2: an airborne collision avoidance system built primarily for passenger airplanes. There are two problems with that system, as it relates to small airplanes. First, it may cost more than some of the airplanes themselves. A used Cessna 172, for instance, costs around $25,000, while one of these commercial collision avoidance systems can cost north of $50,000. Second, these existing systems were built primarily with commercial jets in mind. So, if you were to take something that was designed for an extremely fast Boeing 747, let’s say, and then you were going to put that on a relatively slow single engine piston aircraft that flies at 120 miles per hour, you are not going to get very good results.

At that time, I was taking a class on machine learning at Boston University, and I asked my professor if she would be okay with me working on this idea as my research project for the class. I wanted to build something that could predict in real time how the airplanes around me were going to move in, let’s say, the next ‘n’ minutes, and then a system would alert me-the pilot-and tell me what to do so I could minimize the odds of a collision. So, for example, if there is an airplane that is flying 200 feet below me, the system would alert me and say watch out there is an airplane 200 feet below you. You need to turn 20 degrees to the right and climb 300 feet.

GLTR: Wow, that is an amazing story. What rules must pilots follow while flying?

Tarapore: In visual meteorological conditions (VMC) or clear weather, all pilots—including those flying commercial airliners—are legally required to see and avoid each other. If I were to fly over Boston under visual flight rules (VFR), for example, the pilots of all airplanes around me,  big or small, are legally responsible for looking out the window and avoiding each other.

At non-towered airports (airports without air traffic control) —which most airports in the United States and the world are—pilots are not even required to use the radio! Of course, the Federal Aviation Administration (FAA) encourages pilots to broadcast their intentions, but it is entirely legal to fly solely by seeing and avoiding other aircraft in certain types of airspace.

GLTR: Okay, so how exactly does your technology work to solve this problem?

Tarapore: So, there is a technology called automatic dependent surveillance broadcast (ADS-B), which a lot of modern airplanes use to determine and periodically broadcast their position. Since these positional data points are broadcast over radio, anyone with the right receiver can obtain and interpret them. The ACAS unit samples and compiles this information along with the user’s own aircraft position in real time to identify potential collision hazards.

In simpler terms, if I know where an aircraft is and how fast it’s moving relative to my own position and velocity, all I need to do now is predict where the other aircraft might go in the next few seconds and chart a course away from it to avoid a potential collision.

So, that was the idea. If this information is broadcast publicly, can we take that information and turn it into something useful? That’s what we worked on, and it turns out, yes: all that is required was a little bit of clever mathematics and a lot of very clever systems engineering. You actually can build something that is extremely adept at being able to predict how these airplanes are going to move and, what’s more, you can even tell a pilot how to avoid that collision.

GLTR: So, does a unit of your technology just include the software, or is there some sort of radar hardware also included?

Tarapore: The ACAS device is a portable transceiver that builds on top of the open source Stratux device to receive ADS-B and traffic information service-broadcast (TIS-B) data. It uses this data to identify potential collision hazards and ways to avoid those collisions. It then sends this information over WiFi (with Bluetooth Low Energy as a backup) to iPads running the Foreflight app, which a lot of general aviation pilots use to navigate and fly.

On ForeFlight, the pilot’s own airplane is displayed as a blue dot on an aeronautical map (similar to the dot identifying the car on Google Maps). On this map, the ACAS device superimposes collision threats as similar-sized dots varying in color from yellow to red, depending on the severity of the situation. When ACAS sees a red dot (a severe collision threat), it suggests evasive maneuvers to the pilot to eliminate the possibility of a collision.

It is worth noting that ACAS’s hardware is by no means unique. ACAS’s novelty lies in its ability to understand the odds of a collision and recommend an effective avoidance strategy to the pilot. The algorithm itself will eventually be published in a peer-reviewed journal and will remain open source.

GLTR: Okay, so you have this idea, now you need to implement it. What was your journey as a young tech entrepreneur like?

Tarapore: I’m a good engineer, but I’ll be honest I was not a good entrepreneur. Earlier in the life of my company I had an experienced entrepreneur working with me. He had dealt with raising funding and building a company. When we first incorporated, he took on the role of CEO and helped mentor me, eventually handing me the reins.

With a minimally viable product (MVP) developed and our idea thoroughly validated, we needed funding and began looking for investors. We eventually secured venture funding that valued ACAS at a substantial figure. Our investors have been exceptional mentors and have guided us every step of the way, particularly when dealing with the justifiable regulatory burden involved in augmenting avionics.

GLTR: Were there any legal issues that you faced?

Tarapore: Two legal issues spring to mind, one specific to technology and one specific to aviation. As a college student hacking together a prototypical weekend project, I wouldn’t spend too much time worrying about the specifics of a software’s license. When writing proprietary software for a company, we have to thoroughly understand the terms of the licenses of the various software we use. Open-source licenses without explicit patent grants, for instance, can present expensive legal challenges if we are not careful, and that is not something I knew back in college.

With regard to aviation, broadly speaking, while it is natural to assume that commercial airlines are significantly more regulated than general aviation operations, I did not realize just how stringent the rules are. While validating the idea for ACAS, I learned that it would take years before it could make it to the cockpit of an airliner. This was a bit disappointing, particularly considering that ACAS’s algorithm significantly improves on TCAS. Once our research is formally validated, I hope we can kickstart the regulatory process to request approval for testing this on airliners.

GLTR: Where does ACAS go from here?

Tarapore: Funding has allowed us to recruit and retain experts who are helping turn a prototypical idea into a robust, reliable, and user-friendly device.

We are currently in the process of scientifically validating our algorithm, through approval of a research paper we wrote explaining our technology. When comparing risk ratios, preliminary results show that our ACAS system is about 80 percent more effective than the “see-and-avoid” or TCAS methods of  avoiding collisions. If that holds after the final version of the paper is released, we plan on shipping the first 1,200 units by the end of this year.

GLTR: Great, thank you for taking the time today to speak to me about your startup. Before we go, is there any advice you would offer to someone thinking of starting a technology-based company?

Tarapore:  I tend to come up with a lot of ideas and the biggest thing that powers that is just curiosity.

One of the things I do think that anyone can do —and probably should strive to do —is just ask questions. If you’re walking outside, just ask questions about why things work like they do and that can lead to some pretty remarkable things. So, my advice to anyone is just be a little more curious than you already are.

Dharmesh Tarapore is the CEO and founder of ACAS Technologies, and also teaches computer science practicums part-time at Boston University. He has extensive software engineering experience and particularly enjoys using machine learning to obtain actionable insights from vast datasets. His software engineering career began as a freelance web developer and has now grown to include a sizable repertoire of several varieties of programming. Tarapore graduated from Boston University in 2018 with a Bachelor’s degree in Computer Science.