The Art of Building Robots in Silicon Valley - Part I

Apr, 20, 2024

Long before tech giants like Google, Yahoo, Microsoft, and Amazon came into the picture, robots were already making their mark. In 1949, an American-born British neurophysiologist and inventor, William Grey Walter, built the first autonomous robot in Bristol, England. His invention? Small, battery-powered machines that could find their way to a light source using sensors and software. Fast forward to the mid-1950s, and George C. Devol from Louisville, Kentucky, introduced the world to "Unimate," the first reprogrammable robotic arm. Capable of lifting 75-pound loads, Unimate played a pivotal role in the transformation of the U.S. auto industry but was missing a key ingredient that we hear and see a lot today - Artificial Intelligence (AI).

In 1972, engineers at the Stanford Research Institute developed the first AI-powered robot. This innovative machine could navigate its environment, plan and execute tasks, such as moving from point A to B and transporting boxes, all autonomously thanks to its advanced sensor array. The robotics revolution didn't stop there. In 2002, the Roomba vacuum bot made its debut, followed by Kiva's warehouse robots in 2003, Starship Robots' delivery bots in 2014, and Cobalt Robotics' security robots in 2016. While a number of research institutions and startups from Silicon Valley played a key role in building the next generation of robots, have you ever wondered what it takes to build a robot and to ensure it does the job we intended it to do in the real world?

I interviewed with the founders of a small robotics startup in Silicon Valley around 2018 and was fascinated by their vision – to make this world a safer and secure place. Subsequently, I decided to join this startup - Cobalt Robotics. Our project focused on developing indoor robots designed to automate routine tasks while providing safety and security services to people, properties, and places through cutting-edge robotics and AI systems. Cobalt robots are capable of patrolling office spaces and warehouses, complementing existing safety measures by providing real-time situational awareness. These robots were equipped with advanced sensors like LIDAR, 3D vision cameras, ultrasonic sensors, infrared cameras, and more, making them fully autonomous.

As exciting this venture might sound to most of us it wasn’t without its challenges. We are talking about a product that demands hardware, software, autonomous mobility, and continuous access to internet, all with mission-critical robustness. A product like this is extremely hard to build, launch, operate, and most importantly scale! My job was three fold - ensure our product doesn’t break in the middle of demos, our fleet of robots deployed across the world is fully active 24*7, ensure each new launch that we do is successful - no major incidents or failures at least in the first 2 weeks post launch and ideally never. We will now cover each of these scenarios one by one.

Demo days are exciting and challenging at the same time not because you would need to drive your robots through the tech arteries of Silicon Valley, Texas and Massachusetts but because each demo meant an opportunity to create value for our prospect, investor and employees and last but not the least an opportunity to learn how to not fail again. I encountered a decent amount of failures early on from robot dying in the middle of a demo due to indoor internet blind spots, a last-minute software release that an engineer accidentally released on a robot that was actively being demoed, a sharp ray of sunlight that hit the sensors of the robot so hard that they wouldn’t function as intended - basically all sorts of challenges you can imagine from a mobile 120 lbs device that’s intended to function autonomously in an indoor space. Over the period of next few months, each of these lessons were iteratively shared internally with the team and woven into our product operations protocol. 

Then came the post-sale launch periods. These are thrilling yet daunting times when a client decides to trial our service for six months or a year before fully committing. Think of demos as short sprints, while these phases are the long marathons that continue for at least 6 months and in some cases for years. Now, imagine the challenges we faced on a demo day, but extended over 180 days, with the expectation to hit every performance mark. This phase is crucial for proving our worth and maximizing value for our clients.

Getting clients to test a beta version of our service was a tough task, making early adopters key to our success.  This meant being agile enough as a startup to tailor the service to the needs of the client, even if it involved training a robot to navigate safely through an elevator, working with a telecom vendor in Australia, Japan, or the UK to get your robot to work on global telecom systems, finding the right shipping partner who could take care of customs and shipment, training people to fly overnight across the world to demo and deploy your service, and last but not least, ensuring the robot’s navigation mechanism is fail proof and safe.

Yes, one of the hardest parts of building autonomous robots is to ensure they don’t run into people. This one thing is easier said than done because unfortunately, all the sensors that we use today in robotics have certain limitations and failure scenarios. From sun rays to extreme temperature and viewing angles to the surface of the ground and its slope, and more, there are a lot of variables in this safety equation that can become extremely hard to optimize for safety and security with a focus on delivering real-time services.

Knowing this, the team from the get-go introduced the human-in-the-loop concept. The goal here is that while the robot will navigate autonomously, but for scenarios which we know it may not be able to handle with extreme caution, the robot would raise an alert and ask its navigator to help it maneuver. Doing so the founders and the team ensured, we delivered a service that was safe, reliable, secure and effective. 

While sensor limitations, mission-critical robustness, and scalability of product that is both hardware and software-heavy were some of the challenges we encountered on the technological front, there were others that were more humane.  We as humans aren’t accustomed to interacting with robots so it is completely natural to not have a taste for them.  For instance,  if you are a Fortune 100 company with 1000 employees working on your campus, and if we introduced a safety and security robot to help your employees, not all employees would be willing to opt for such a service and chances are most might either mock the concept or would be skeptical of such a service.

Thus in robotics, while many think technology and operational barriers are the biggest challenges, I would argue that psychological barriers are probably the hardest to overcome. We all know we would much rather have a robot equipped with a deterrent tool and video-chat functionality to negotiate with an armed man than risk the life of a security guard if we had the option as our primary mode of resolution. However, the reality is that we humans have yet to develop trust and a taste for robots, and that's largely due to our evolutionary cycle and, partly, Hollywood - sorry, Terminator and Transformer fans.  It reflects a broader challenge of familiarizing society with the benefits and reliability of robotic assistance.

Through our robotic venture, we learned that these barriers can be navigated early on by conducting more pre-launch lunch-and-learn sessions with clients and employees, being transparent about how the robot works and why it does what it does, demonstrating multiple high-risk scenarios in a controlled manner, and sharing more case studies of clients who have benefited from these services immensely. Setting the right expectations and being transparent about what works, and what doesn’t with all your stakeholders is key for building trust and scaling your startup. 

If you or someone you know is contemplating a career in robotics or plans to start a robotic startups, remember, robotic service is effective as long as you know exactly how it works, you have studied 99% of all failure scenarios, and have guardrails in place for the remaining 1%, and last but not least, you are transparent about the successes and failures of your service with your investors, clients, and employees.

That’s what the founders did at this startup, and they are some of the most inspirational people I have met to date.

Until next time - Hello Robot!

Cheers,

Prathamesh 

Disclaimer: This blog is for educational purposes only and does not constitute financial, business, or legal advice. The experiences shared are based on past events. All opinions expressed are those of the author and do not represent the views of any mentioned companies. Readers are solely responsible for conducting their own due diligence and should seek professional legal or financial advice tailored to their specific circumstances. The author and publisher make no representations or warranties regarding the accuracy of the content and expressly disclaim any liability for decisions made or actions taken based on this blog.