Historically there were two main types of robots: industrial and toy/home. Part of the reason for this was that robots used to be lumbering oafs that couldn’t see or sense their world and would run into anything in their way. In industrial settings, these robots were and are extremely dangerous and are literally caged off from humans.
The other types of robots were toys or home robots that because they might bump into something had to be lightweight and slow. Robot toys were often covered in padding or fake fur. The most successful home robot to date has been the Roomba, and it can only move in one direction with a redundant bump sensor on the front to make sure that it can’t hurt anything.
But now the dual advances in sensors and connectivity are driving the emergence of cobots designed to interact with us.
One of our portfolio companies, Marble, just launched their robotic last mile delivery service. They currently have human operators walking with the robots, but eventually the robots will take over the deliveries. When the robot encounters a situation where it gets stuck or doesn’t know what to do, operators at Marble will be able to dial into the robot to take over and manually drive.
Historically, when you deployed a robot, it was all or nothing. The robot had to work 99.99% of the time or it was no good. The world is a complex place and can be confusing for robots. Now we can program them to stop and call for help for that .01% of cases, which means they can be out in the world far earlier than they could before.
However, the major reasons behind the cobot emergence are an improvement in performance and a reduction in the cost of sensors. Sensors like lidars coupled with cheap cameras give a cobot the ability to see and avoid items in their environment. The equipment with the ability to detect, process, and react is now cheap and small enough to embed onboard the robot. Lidars that cost $50,000 today were millions of dollars just a few years ago. Before that, there wasn’t a lidar in the world that could perform as well as the Velodynes do today.
The final major hurdle to cross is to have a cobot that can bump something and stop before it harms that thing or itself. The first robot to really do this was the Rethink Robotics’ Baxter. Still slow and cautious, Baxter’s embedded sensors and the design of its arms let it work side-by-side with people.
Over the next few years, we will see robots deployed that are fast and powerful when people aren’t nearby but that slow down and take caution when people approach. We are excited for this. As a result, a key part of our focus on robotics startups is cobots and we’re actively looking for robots that can be broadly deployed in the real world.
If that’s something you’re working on, please reach out! I’m firstname.lastname@example.org.