Kevin Keogh, Starsky’s System Validation Lead: “It’s okay to not have the answer, we’ll solve this together”
Recently, Starsky Robotics made history by performing the first ever unmanned run on a public highway. Kevin Keogh, Starsky’s System Validation Lead, has contributed greatly to this success. We talked with Kevin about what made it possible for Starsky to get ahead of the competition, the role of System Validation, and why a good company culture is essential for making a safe product.
Kevin, tell me the story of how you joined Starsky. You were living in Ireland back then, weren’t you?
I’m from Ireland, but was living in the UK at the time. I was working at Jaguar Land Rover on some AI features for their cars and I liked the job very much. I was doing really cool things day in and day out. But the midlands of the UK was not the most exciting place to live. Both myself and my wife Rebecca had very long commutes. We worked hard, and at the end of the day, we would find ourselves in a sleepy town without much going on.
It was a crazy right-place right-time situation that put me in touch with Starsky. One morning I called my best friend Matt who was in San Francisco at the time, and I mentioned to him that I was thinking about making a move.
That evening Matt met Stefan at a start-up event in SF. Stefan told Matt about Starsky and his vision for autonomous trucks, Matt told Stefan that I worked in a similar space and was thinking about a move, Matt put us in contact. Back then, in the early days of Starsky, Stefan and Kartik were taking the very first steps towards building a self-driving truck company. We did some interviews over Skype and I was hooked as soon as Stefan outlined his thoughts behind why he wanted to build Starsky. The business case, and combination of teleoperation and autonomy made so much sense to me.
You are the System Validation Lead at Starsky. Can you please explain a little bit more about what it means and what you and your team are working on?
An autonomous vehicle is a highly complex product and, because of its use case, it’s extremely safety-critical. Given these factors, we need a way to know that we are building the right features into our system and also that those features are performing the job they are designed to do. System Validation performs these tasks and is an essential part of our process towards functional safety.
When building a complex product like ours, we need to enable teams to work together in effective ways. For example, in the robotic system, planning and controls modules will depend heavily on the perception module to provide them accurate information from the environment around the vehicle. It needs to be very clear exactly what perception will provide to the other teams. In a way, each team agrees to perform certain functions, these agreements are made in a form that we call engineering requirements. System validation helps with the definition and design of these requirements, then also provides an effective framework to validate that each requirement is met in our system.
Starsky has been first in many things. Last year, Starsky did its first unmanned run in Florida and became the first company to test an empty cabin for autonomous trucks. Starsky’s truck set a record for the fastest unmanned road-legal vehicle. Finally, Starsky just drove a truck with no human being on-board on a public highway. What, in your opinion, made it possible for the team to get ahead of the competition?
We have had the unmanned test as a goal at Starsky for quite awhile. Since very early in our development, we knew it was going to be a key milestone on the way to productization. It takes a lot of time to develop and validate a system like ours. What has been a huge advantage to us at Starsky is an extremely dedicated team of teleop and safety drivers. We have a team of highly experienced drivers, many of these people have greater than 20 years of experience driving trucks. They have a wealth of invaluable domain knowledge. In Starsky’s early days, their input on vehicle dynamics and truck hardware saved us valuable time on autonomous development, allowing us to hit key early milestones.
Also, I think what made it possible, which is also one of the main reasons why I joined Starsky Robotics originally, was an extreme focus on our business case from day one. The thing that really made me want to join this company was the approach to deployment of autonomous vehicles. The focus on semi-trucks as a vehicle that spends 99% of its time on the highway makes a lot of sense for autonomization. Long highway stretches are the perfect environment in which to actually deploy an autonomous vehicle. The other 1% of the time, or what we call the first and last mile, is where the AV industry is investing billions of dollars looking for a solution.
Our approach of teleoperation and supervised autonomy for these sections has been the key factor that has allowed us to move forward at such a pace. Our combination of highway autonomous, supervised autonomy, and teleoperation is a major factor in our success.
From the system validation standpoint, what does it take to get to the point where you can safely take a person out from behind the wheel on a public highway? What were you feeling on the day of the test, watching the truck with no one in it driving in live traffic?
It takes a lot to get to safety confidence. A fully defined safety case, combined with an Operational Design Domain that exactly fits the vehicle performance and road environment is required. Without these elements, you cannot make the structured approach that is needed. We need to know that the vehicle is capable of performing its driving task without errors. We also need to know that in the unlikely event of an error occurring, the vehicle can react safely with a minimal risk condition.
Running up to our unmanned test, I always thought I was going to be nervous. This was a huge goal for us, and nobody in the industry has done it before. But to be honest, on the day of the run, it felt really natural, our test process didn’t allow for nerves to enter the equation. We had our full plan laid out months beforehand, we followed it exactly step by step. The vehicle passed all our tests, and all our personnel were informed and prepared for every scenario. We really did our job on the foundation steps. The morning of the tests, we ran our final pre-flight checklist, everything was green, and that was it. The test was go.
Then the test ran exactly the way it was supposed to. I was sitting in a car behind the truck and forgetting that there was no human being in the vehicle. It hits you when you’re watching: “Oh yeah, there’s nobody in the truck.”
What is it that you like most about your job?
At Starsky, I touch on all elements of our robotic development, and I really enjoy having this broad system level exposure. When I joined Starsky, I was the third employee and the first robotics perception engineer. So, I built up the initial perception stack and some of our first control algorithms. Because it was so early-on back then, at the very start of our autonomous development, I was involved in everything, from our full robotic stack, to our computer platforms, even to some hardware. I was hands on in every part.
This involvement at the full system level has allowed me to form a career path that really fits my skill set and interests. My robotics and physics background as well as my automotive experience has guided me to this systems and validation role in which I can ensure our safe and effective autonomous system development. I really love it.
The company is growing very fast. Starsky almost doubled in size since last year. There are new people joining every week. But what’s interesting is that despite that crazy growth, people who work here feel like they are part of a big family. What helps keep the company’s culture tight-knit?
I think at Starsky, our culture was started the right way very early on by our founders. When you’re a small start-up trying to do a crazy thing like build an autonomous vehicle, you tend to spend a lot of time with each other and it’s never going to work if it’s very hectic or if people aren’t getting on well together. So, from the get-go, there was always good camaraderie, teamwork, and support.
We are building an extremely difficult product that no one has built before. So, when we hire someone, we don’t expect them to know everything. We don’t expect them to have all the answers coming in the door. But what is really important is that anyone who joins us is able to work as a cohesive part of our team. Together, we work to build the best product, and make the best decisions. It’s something that we’ve focused on while hiring. We understand that we have this good collaborative culture and we’re working hard to keep it.
If there’s anything you are struggling with, you should be comfortable asking questions. Sometimes when you are trying to figure something out, it might take you 20 hours to research, when actually a two-minute conversation with another one of our engineers will lead you to the right solution. This is a valuable conversation to have. Also vice versa with things that you personally are an expert in, you should be able to help other team members, and share your knowledge to bring the whole team forward.
I think that has been a key element of our success, our friendly collaborative environment where, it’s okay to not have the answer, we’ll solve this together.
Can you share an interesting fact about yourself that you think is important to know about you?
I think what’s interesting is that I had a chance to get familiar with trucks long before Starsky, when I was a kid. I grew up on a potato farm in Ireland. My family are all farmers back home, and growing up, a lot of life revolved around the family business. I wasn’t very big into farming as a kid, but from a young age, I had a lot of experience with the heavy machinery like tractors and trucks.
A really fond childhood memory revolves around some of our truck deliveries. Basically, every morning at 3am, a truck would leave from the yard and deliver potatoes into the corporation fruit market in Dublin. We had a great truck driver called Gerry Fay who would take that shift, but on the weekends my dad would cover it to give Gerry a break. In the summer holidays, when I wasn’t in school, I used to get up early to go with my dad. It was super exciting going to Dublin city in the big truck in the early morning. I remember being in the passenger seat leaning forward with excitement and then my dad shouting at me to sit back because I was blocking the side mirrors. It was a lot of fun. Who would have thought that I was actually the kid who would end up automating these gigantic vehicles?
***
If you want to join the Starsky team and help us get unmanned trucks on the road, please apply here.
