What Is a Workhorse Robot?

Toro Newsroom

Over the years, robots have been sliced and diced into many categories: industrial, personal/consumer, autonomous vehicles (AVs), drones, educational, entertainment, medical, humanoid, underwater … and the list goes on. While people imagine the future having robots in many aspects of life, historically, they have had most traction in manufacturing logistics. In this area, these are the common categories of robots:

• Industrial robot: This robot is generally fixed to the floor of manufacturing plants or warehouses. A classic example is the manufacturing arm, in which robots were manually preprogrammed to perform repetitive tasks with no sensing of the environment around them. With recent innovations in computer vision and haptic sensors, industrial robots are becoming smarter.
• Automated guided vehicle (AGV): Like the industrial robot, the AGV also commonly operated on manufacturing and warehouse floors with little embedded intelligence. It requires some type of guidance system in a perfectly controlled environment — for example, a painted line that it follows or a buried wire.
• Autonomous mobile robot (AMR): This newer, smarter vehicle operates in similar environments as the AGV. The AMR uses sensors that allow it to better perceive its surrounding environment, thereby allowing the AMR to operate in more dynamic environments and better collaborate with humans.

I don’t know about you, but I find these categories pretty confusing. Why is an industrial robot always thought of as fixed versus being mobile? And now that everything from your toaster to your car is becoming smarter, AMR is an extremely vague definition for a robot.

It might help to take a look at the history of robots to understand how we got here.

While the term “robot” has been around for the last 100 years (the term was first used in a 1921 play called “Rossum’s Universal Robots”), it was not until the 1980s that robots moved from science fiction to actually being used in industry. This is when the first wave of robot arms began being used in manufacturing plants. This kicked off the category of industrial robots: manually programmed robots that had few, if any, sensors for environmental input. At the time, industrial robots had little intelligence and mindlessly repeated the actions they were programmed to accomplish. Interestingly, at the same time, university research was started in leveraging artificial intelligence (AI) to enable self-driving cars. In 1986, Carnegie Mellon University launched NavLab, one of the first “cars” designed to be driven by a computer. It actually was a large van packed with five racks of computers, including a Warp supercomputer.

It was at this time that AGVs also showed up on manufacturing floors — the mobile counterpart to the fixed industrial robot. Other than these two industrial use cases, for the next 20 years, robots stayed in the university labs. One small exception to this was toy robots, but they represented a small percentage of the robot market.

Through the 1990s and early 2000s, there were incremental improvements to industrial robots and AGVs but no significant change in how they worked. In the last decade, AGVs started to be replaced with AMRs — named to indicate they were mobile — and instead of simply being guided, they had the ability to sense and autonomously interact with more dynamic environments compared to their simple-minded AGV predecessors. So, as you can see, much of the context around the differences between industrial robots, AGVs and AMRs is based on the history of this equipment being used in manufacturing logistics.

Now that AI and computer-vision-driven perception is being used to guide all sorts of equipment and vehicles — mobile robots in warehouses, delivery robots on sidewalks, robot boats, self-driving cars, autonomous drones, etc. — terms like AV, primarily used for cars, and AMR, primarily used for smart mobile logistics robots, are sloppy category names. I instead propose that we use a broader label — “smart things” — to define all vehicles, equipment or other things that leverage sensors, AI and other processing to perceive and interact with the world around them. Smart things include robots and can be nicely mapped across these two dimensions: fixed versus mobile and work/task oriented versus social/entertainment oriented. See the following diagram.

In this diagram, the green items are more consumer oriented, while the other colors are more business oriented. The green items are all terms that are in use today. The distinction between autonomous trucks and autonomous cars should be self-explanatory. Industrial robots are the descendants of the dumb, fixed industrial robots of the 1980s, which will continue to incorporate evermore increasing sensing and intelligence. Workhorse robots, in red, is the newest term.

A workhorse robot is a fully autonomous, skill-specific vehicle that leverages much of the AI and computer vision technology historically driven by self-driving car research. These robots are the next-generation AMRs, built to cover tasks that extend way beyond the traditional logistics space that AMRs previously operated in. I predict some of the most interesting business opportunities over the next several years will be in workhorse robots. As the cost of perception technology shrinks enough to support mass adoption, the economics of these robots will support all sorts of tasks never before possible to machine automate

A workhorse robot is a fully autonomous, skill-specific vehicle that leverages much of the AI and computer vision technology historically driven by self-driving car research.

I believe that all the categories of robots and other smart things can fall into one of the bubbles in the diagram, but let’s take a deeper dive into the types of solutions that can fall under workhorse robots.

• Agriculture: GPS-guided tractors have been available for years; now, robots can help with more advanced tasks in the field from weeding to plant health inspection.
• Construction: Robots can aid in repetitive but somewhat complex tasks from digging and moving materials around outside to building walls and structures for real estate construction.
• Retail/customer service: Retail solutions include tasks such as stock inspection, restocking and even helping customers as mobile information kiosks.
• Cleaning: This category has exploded in popularity with the COVID-19 pandemic. Robots disinfect surfaces in plants and retail stores or sweep or mop floors, parking lots and sidewalks.
• Inspection, surveying and security: These robots can inspect and survey areas that are dangerous or difficult for humans to access. Robots lead to more accuracy because they can repeat tasks more reliably, operate for long periods of time without breaks and provide more comprehensive coverage.
• Logistics and delivery: More intelligent workhorse robots can help with logistics and delivery tasks in much more dynamic environments than their AGV predecessors. As the cost and complexity of operating these robots decrease, the more efficiencies can be gained.
• Medical: These robots perform tasks in hospitals and other health care facilities. They can perform delivery tasks, administer medication and monitor patients’ status. They can operate longer with more precision and fewer mistakes than humans doing similar tasks. They also can operate with less risk of disease or other infection transmissions.
• Military and field: These robots can be used in combat or military reconnaissance or be sent to locations difficult for humans to get to for exploration tasks (for example, space and other planets, deep-ocean exploration or extremely remote locations).

While these vertical solutions are primarily described from a business viewpoint — there are also consumer versions of workhorse robots. For example, consumer agriculture solutions include lawn mowing; consumer cleaning includes robot vacuums and window washers.

To conclude, my hope is defining this new space of smart things, in addition to the category and sub-categories of workhorse robots, helps to provide a better framework for talking about the progression of robots and smart equipment and how they will impact our lives. It’s time to move beyond the more ambiguous terminology developed from the historical robot evolution in the manufacturing logistics space.

Leo is the chief innovation officer at DataTribe, a cybersecurity and AI/ML startup foundry: “A place where the world’s best come to build dominant companies.”