Back
May 15, 2024

Less-than-perfect robots deliver huge amounts of value

Ira Renfrew

Chief Product Officer

Science fiction has taught us to expect “perfect” robots – robots that can handle complex tasks, learn new information, and communicate with us as well as, or even better than, humans – think Baymax, Chappie, IG-11, and many more. Even the simpler sci-fi robots, such as Wall-E and R2D2, are highly proficient. They handle highly variable tasks and dynamic environments with ease. This has led robot developers and the general public alike to expect robots to be highly capable.

Spoiler alert: We’re nowhere near sci-fi capabilities with the robots we’re building today.

We can think of the world of proven robot applications in a simple 2×2 matrix – on one axis, tasks can be standardized or variable. On the other axis, the environment can be static or dynamic. The combination of these axes is a measure of how much situational diversity the robot encounters.

Historically, robots have been deployed to do standardized tasks in static environments – in these cases, the situations the robot faces are uniform. Good examples of this are production lines, where robots repeat the same stitch, weld, or bolt tasks repeatedly. These robots rarely need human help.

Robots, including autonomous vehicles (AVs), have more recently been taking on increasingly variable tasks in dynamic environments, such as homes, sidewalks, and warehouses – the situations the robot faces are increasingly diverse. However, despite the recent round of AI and humanoid robotics hype, there’s still a limit to the situational diversity that robots can handle, a limit that they currently struggle to pass.

This typically results in two buckets of work: 1) the more uniform situations with less variable tasks and less dynamic environments that robots can handle, and 2) the diverse situations with higher task variability and/or highly dynamic environments that they can’t handle. Humans, who adapt quickly with flexible mental and physical capabilities, continue to be ideally suited to dealing with this second bucket. Rather than trying to build robots that try to handle the second bucket, robot developers should focus on designing for the first bucket and rely on humans for the second.

But does this make business sense? It depends. For a robotic product to make business sense, the first bucket needs to be the large majority, or baseload, of the tasks to justify the time and cost of developing and deploying the robotic system. The second bucket must be no more than a “long tail” of diverse situations that, at least initially, can be given to humans to address. If this condition is met, the path is clear to developing a robotic product that delivers immediate value to customers.

I first encountered this situation at iRobot with the Roomba vacuum cleaning robots. Homes are dynamic environments, and cleaning them involves vacuuming open floor spaces, under furniture, in corners and tight spaces, and on stairs. We could have tried to develop a robot to do all this, but it would have taken years – along with a hefty price tag – to reach consumers.

Instead, iRobot realized that if the robot could vacuum the open floor spaces and under furniture (the majority of the work) for a reasonable price, then customers would be happy to finish the job by vacuuming the tight spaces and the stairs (the long tail of work) that the robot couldn’t reach. We reduced the situational diversity by limiting task variation. Millions of happy customers validated that approach and iRobot created an entirely new product category. The robot vacuum didn’t have to be perfect and do everything, it had to do enough.

We enthusiastically take this approach to product development at Outrider as we automate moving semi-trailers on distribution yards. Distribution yards are a headache in the supply chain — inefficient, costly, and unsafe – with little historical innovation. Trailers are moved around a facility multiple times before they go back out on the highway, and the basics of each move are relatively standard. But because the environment is dynamic and each trailer is a little different, these trailer moves split into a baseload of relatively standardized moves and a long tail of trickier situations. For example, a tight corner in the yard might require the truck driver to jackknife their tractor to back a trailer into a spot – an impressive feat to watch but a hard one for a robot to pull off!

So, instead of trying to build the “perfect” sci-fi yard truck robot, we’re building a system that handles the baseload of moves around the distribution yard – 80% to 90% of the moves – while leaving the long tail of edge cases for humans. Our robotic system hitches, backs, connects and disconnects brake lines, safely maneuvers around obstacles, manages trailer inventory, and communicates with the dock doors. It does this for the vast majority of trailer moves on any given site on any given day or night.

For the long tail of challenging situations, we rely on humans. Our system asks on-site or remote staff for additional information, such as the location of a trailer’s gladhand (the connection point for the brake lines), or it requests assistance in removing an obstacle blocking the AV’s path, like an over-the-road truck idling in front of our destination spot. Our system can also send the trickiest moves and tasks to humans, such as connecting to damaged gladhands.

Over time, our system will evolve to handle more and more of these diverse, long-tail situations. But instead of waiting for that day, our system can offer value to customers now by focusing on the baseload of moves. Because we can cut off the long tail of difficult situations, Outrider can deliver value to our customers sooner.

Outrider is reinventing supply chain logistics and transportation by seamlessly integrating robots — 20,000-pound autonomous yard trucks — into yard operations. We’re the first to develop these enormous robots, which do the baseload of the work and only occasionally call for human assistance.

Like many, I look forward to the day that sci-fi-like robots become part of our reality. But in the meantime, I gladly focus on shipping robots that create value now instead of waiting for the perfect robot.

Ira Renfrew, Chief Product Officer

Ira Renfrew is the Chief Product Officer at Outrider, the leader in autonomous yard operations for logistics hubs. He is committed to improving the quality of everyday human experience through the design and deployment of robotic solutions. Under his product leadership, Outrider will scale the first commercial robotic system that improves work efficiency, safety, and sustainability in global distribution yards.

You might also like

February 22, 2024

It’s 2024, and we are shipping our commercial system later this year

Later this year, the world’s largest F500 companies will begin autonomously and driverlessly ... READ MORE

November 14, 2023

Navigating challenges and seizing opportunities as a female engineer

A career in the tech industry has always been in the cards for ... READ MORE

November 09, 2023

Alleviating the Bullwhip Effect: The Promise of Autonomous Logistics

The year is 2020. The COVID pandemic is in full swing, and there ... READ MORE