How Baxter the Robot Works

Rodney Brooks, the man behind Baxter, is also responsible for many of the robots that already exist in homes today. If a Roomba takes care of your light vacuuming duties, you can thank Mr. Brooks.
Rodney Brooks, the man behind Baxter, is also responsible for many of the robots that already exist in homes today. If a Roomba takes care of your light vacuuming duties, you can thank Mr. Brooks.
Image courtesy of Rethink Robotics

Rodney Brooks has been one busy roboticist. He joined the faculty of the Massachusetts Institute of Technology (MIT) in 1984 and eventually became the Director of their Computer Science and Artificial Intelligence Laboratory. While still there in 1990, he co-founded iRobot with two of his students in an effort to get back to more hands-on robotics work. That company eventually brought Roomba and other domestic cleaning robots into millions of homes worldwide, edging out the scant competition with a much better price point.

After that success, Brooks decided he wanted to revolutionize another industry with robotics, so he stepped down from iRobot and MIT and founded a new company called Rethink Robotics (formerly Heartland Robotics). Rethink seeks to put more robots into the manufacturing industry. Manufacturers already use heavy industrial robots in droves. Industrial robots are moving machines that can be programmed to manipulate materials and tools to perform various manufacturing tasks. They are used to automate the building of cars and production of a host of other goods. As a rule, they are extremely expensive to purchase, program and maintain, and they are therefore out of the hands of anyone but very large manufacturers. Such robots are also very precise and fast, but do not adapt easily to new surroundings and cannot work in close proximity with humans without risk to life and limb.

Brooks and his company therefore set out to make a cheap and adaptable robot that could be purchased by smaller manufacturers and used to automate mundane tasks that are currently being performed by people. And their work has come to fruition with the release of Baxter, a $22,000 robot that can be safely placed into a human work environment and easily trained by a non-technical person to perform a variety of repetitive tasks.

What physical components make Baxter tick?

Baxter’s expressions let nearby workers know what’s going on with the robot’s processes, from stand-by to concentration to a sad resignation when a task isn’t going well.
Baxter’s expressions let nearby workers know what’s going on with the robot’s processes, from stand-by to concentration to a sad resignation when a task isn’t going well.
Image courtesy of Rethink Robotics

Baxter is a somewhat humanoid robot, but it wasn't made this way just for the sake of appealing to humans. All its parts serve a purpose. It has a computer screen head somewhat reminiscent of a tablet, on which animated cartoony eyes and eyebrows change expression to convey intuitive messages to nearby workers. These expressions include closed eyes when it is on stand-by; neutrality when it is ready to start learning (eyes wide open and eyebrows parallel to eyes), concentration when its in the middle of learning (eyes wide open and eyebrows slanted down toward the center), focus when it is working without any issues (eyes looking downward and eyebrows slanted down toward the center), surprise when a person comes nearby (wide eyes with dilated pupils, eyebrows raised and an orange screen background), confusion when it's having an issue with a task (wide eyes with one eyebrow inverted and both slanted downward toward the outside) and sadness when there is a problem and it has given up trying to work on a task (eyes looking downward with both eyes and with eyebrows inverted and slanted down toward the outside). Baxter's eyes also move in the direction one of its arms is about to move as a warning to anyone working nearby.

On its torso, two arms are mounted that measure 41 inches (104 centimeters) from shoulder to end-effector plate, an area with interchangeable end-effectors, or "hands," that serve various purposes. Baxter comes with two hand types: a vacuum cup and an electric parallel gripper with something like fingers. The cups and fingers can also be swapped out for others, and it is expected that third parties will develop new types of end-effectors. Each arm has seven degrees of freedom, and they can work in tandem or independently of one another, depending upon need. This means you can do things like put Baxter in between two conveyor belts and have it work both. The arms are also compliant, a robotics term meaning that rather than being completely rigid and unable to change course, they can sense and adjust to any obstacles they encounter. If you grab, push or bump into one of its arms, it will give rather than remaining fixed. This compliance is made possible by using series elastic actuators, in which a motor and gearbox control a spring that drives the joints, rather than directly controlling the joints. It's the springs that make each arm less rigid than typical robot arms, and they are also used to measure forces acting on the arms.

Baxter's ability to sense when it comes into contact with something comes in handy for safety reasons. Baxter also has a 360-degree sonar sensor and five cameras. Its software was designed to use these hardware devices so that Baxter can, in effect, feel and see things in its surrounding area and adjust its actions accordingly. When it senses that someone has come close, Baxter displays a surprised expression and slows its movement. When it unexpectedly bumps into something or someone, it stops moving entirely. And if the power is cut for any reason, Baxter's arms slowly relax. Its arms also have back-drivable motors, padded coverings and a lack of pinch points for added safety. There is also an emergency stop button, just in case. And the cameras aren't just for safety. They allow Baxter to detect objects that it needs to manipulate in its work area.

Like many of its robot brethren, Baxter doesn't have legs, so it can't get around on its own. It can either be mounted to one spot or can be placed on an optional wheeled pedestal and rolled wherever it's needed. Baxter stands 37 inches (94 centimeters) tall without the pedestal, and with it, between 70 inches (178 centimeters) and 75 inches (191 centimeters) in height. Its weight is comparable to a human at 165 pounds (75 kilograms) without the pedestal. The pedestal adds 141 pounds (64 kilograms).

Baxter also has input and output (I/O) connections that include an Ethernet jack, a USB type A port and a 15-pin D-sub with PLC-compliant connections for interfacing with other devices. You can plug the robot into the typical 120-volt outlet.

What Can Baxter Do?

Baxter will happily work for hours without needing a bathroom break.
Baxter will happily work for hours without needing a bathroom break.
Image courtesy of Rethink Robotics

In terms of operational capacity, Baxter has a maximum payload of around 5 pounds (2.3 kilograms) and a speed of 2 to 3.3 feet per second (0.6 to 1 meter per second), depending upon payload. It can perform up to 12 pick-and-place operations per minute using both arms. Baxter is rated to operate in temperatures from 32 to 104 degrees Fahrenheit (0 to 40 degrees Celsius).

Baxter was designed to handle a variety of repetitive tasks normally performed by humans on a factory floor. According to Rethink, these tasks include:

  • Loading and unloading of lines, racks, etc.
  • Packing and unpacking items to and from boxes, thermoformed trays or cases
  • Putting products into retail blister packaging
  • Light assembly, such as press-fitting plastic parts into place
  • Operating heat and seal, welding and stamping machines, among others
  • Performing test and sort functions, such as weighing parts and performing visual product inspections to remove defective items
  • Squeezing and crimping tubes
  • Putting caps on jars

To train Baxter, a person physically moves its arms and uses navigator buttons on the arms to make selections from the screen. Say you wanted to program it to pick up and place objects. You would first grab the robot's wrist to let it know it's engaging in training. Then, you would position its end-effector over the item. A camera will center on the item and display it to the screen. You click to confirm that it is the item you wish Baxter to pick up. After Baxter picks it up, you position its arm over the destination, and click to confirm. If the destination is a box, Baxter will pick up the item and place it in the box. Once the training is finished, Baxter will continuously grab and place items based on your prior instructions as long as there are objects for it to grab. Baxter can adjust to item position and orientation changes to a point, so as long as the items stay in the same relative area (for instance, while they are coming down a conveyor belt), Baxter should be OK. If it runs into trouble, its facial expression will change accordingly and a human can intervene.

Baxter has quite a bit of basic built-in functionality now, but there are plans to release regular software updates to improve Baxter's capabilities. The robot's software is based on ROS (Robot Operating System), an open source BSD Unix-based OS which includes a specialized set of drivers, libraries and other tools specifically for programming robots, and OpenCV (Open Source Computer Vision Library), a software library that includes computer vision and machine learning algorithms. A software development kit (SDK) is being released by Rethink sometime in 2013 to allow for future development of more complex capabilities via third-party programming by companies and robot enthusiasts alike. Baxter was designed to be an extensible platform with lots of room for future improvement through hardware and software add-ons.

How Does Baxter Compare to Other Industrial Robots?

Baxter is not able to do the heavy-duty work that robots like these do in an automotive plant in Tianjin, China, but it’s much easier and more affordable for smaller companies to integrate Baxter into their workplaces.
Baxter is not able to do the heavy-duty work that robots like these do in an automotive plant in Tianjin, China, but it’s much easier and more affordable for smaller companies to integrate Baxter into their workplaces.
© ChinaFotoPress/Getty Images

One of the major differences between Baxter and other industrial robots is price. Most industrial robots cost $100,000 or more to purchase, and much more to maintain and operate due to the need for programmers to write the code that controls them [source: Kelly]. They also generally need to have specialized environments built around them to work efficiently and keep humans safe, so you also have to factor in engineering and construction costs. This requires a substantial investment, both short term and long term, possibly totaling hundreds of thousands of dollars.

Baxter, on the other hand, costs $22,000 as a base model, which includes a one-year warranty and a year of software upgrades [source: Rethink Robotics]. You can also get a three-year warranty for an additional fee. Rethink managed to keep the price low by keeping cost in mind during the design process and working with parts suppliers to determine what capabilities their products could bring to Baxter. It also has a lot of plastic parts. Because of the low cost, Baxter should be within the reach of small to medium businesses that currently cannot afford automation via robotics. Rethink estimates that the cost of operation works out to about $4.00 an hour.

And rather than requiring high-level technical staff to spend days, weeks or months programming (via code or the use of a push-button pendant), Baxter requires far less expertise and time to instruct than most industrial robots. It has a somewhat intuitive user interface via the facial expressions and prompts displayed on its screen. A non-technical person can teach it what to do through arm movement and simple button presses, and it can master a new task in half an hour or so. There is also little assembly or setup required. It only takes about an hour to get Baxter up and running once it's taken out of the crate.

There are also huge differences in safety and flexibility. Other typical industrial robots have to be kept in cages or otherwise locked away from humans, lest their rigid, fast and powerful movements injure or kill someone. Their work environments have to be designed around them, rather than them fitting into existing human-occupied environments. Baxter can work alongside humans without risk of injuring them due to its built-in safety features. It can adapt to its surroundings due to a range of sensors, and it has been programmed to have a certain level of common sense. For instance, it knows that if it drops an object, it needs to stop and retrieve another before continuing its motion. And it can adjust accordingly if a conveyor belt speeds up or slows down. Baxter's adaptability and safety allow it to be plopped onto an existing assembly line without too much fuss.

Baxter does have some limitations in that it is not as fast or precise as existing factory robots, and it can't do heavy lifting. It's more adept at the human than the superhuman. This means that rather than replacing existing robots, Baxter could be used to fill new roles so that current manual but mundane tasks could be automated relatively cheaply.

Is Baxter likely to put humans out of work?

Baxter works performing pick-and-place tasks while a human worker supervises.
Baxter works performing pick-and-place tasks while a human worker supervises.
Image courtesy of Rethink Robotics

Lots of our literature reflects our fear that robots will go berserk and kill or enslave us. That's unlikely with workbots like Baxter, which have lots of functionality but lack sentience. Baxter can't "think" for itself. But a more rational fear is that they will put people out of work. Distrust of automation as a bringer of unemployment is not new and has resulted in protest throughout the machine age. One such protest occurred with the introduction of the mechanical loom in France in the late 1700s, which you may have heard about in the movie "Star Trek VI." Textile workers, afraid for their jobs, threw their wooden sabots (shoes) into the looms to stop them, giving us the word sabotage.

Machines have, indeed, taken over many jobs formerly performed by humans. In the 1800s, most jobs were agrarian, but the Industrial Revolution ushered in new technology, and we began mass-producing goods, and even farming, with machinery and automation. Over time, this completely transformed the jobs that most people do, moving us away from farms and into factories and offices.

It wasn't always good news, as sometimes these innovations have caused unemployment, like in coal mining communities that were driven into poverty. They also led a lot of people out of the fresh air and into grueling low-paying factory jobs. But better paying and less physically demanding jobs have also cropped up as a result of mechanization, like those of the technicians needed to operate, program and maintain the machines. Eighty percent of the work in automotive plants is now automated, and this change has made the work less backbreaking and more satisfying than before for remaining car factory workers.

Without machine automation, we wouldn't have been able to mass-produce the tiny parts that have brought us into our modern age of advanced technology. We certainly wouldn't have been able to make the increasingly small and powerful microchips that run our computers, another invention that has greatly changed the employment landscape. Many of us now do things like pay our bills, file our taxes and buy plane tickets via computer, reducing the need for us to deal directly with postal, tax and travel professionals, to name a few. As we begin to handle chores that used to take more human interaction with button clicks, jobs will shift. And there is no end to the things that can be automated. There is even software that writes sports and stock articles for professional news organizations. Full articles like this one could be next.

Baxter could displace some factory line workers. With its current capabilities, Rethink has estimated that Baxter can handle the simple materials handling jobs held by about 800,000 people in the U.S. [source: Freedman (Inc)]. The low price means that the cost of a human worker can be made up in a year or so. Plus, Baxter can work non-stop for longer periods of time. But robots may just shift what jobs people do away from the most menial assembly line tasks into better jobs, like training and maintaining the robots. And there are many tasks that require more thought than Baxter can muster. Robot line workers will also be good for positions that it's hard to find humans to fill, like late shifts, or jobs that cause repetitive stress injuries or expose people to noxious fumes or other dangerous conditions.

Low-cost industrial robots could also reinvigorate the U.S. manufacturing field by being cost-effective enough to compete with lower wage workers overseas, which could lead to more local factories and less outsourcing. The United States actually lost jobs when it was slower to automate than countries like Japan, as was the case with the automotive and electronics industries in the '60s and '70s. It could be that if U.S. manufacturers don't adopt more industrial robots, companies in other countries will do so and gain or retain the ability to produce things more cheaply. Increased access to inexpensive robotics could both free workers from physical drudgery and allow more companies to effectively compete in today's global economy.

Further, we probably shouldn't hold back progress for jobs that won't always be necessary. Computers and robots have led to the ability to perform calculations and tasks with more speed and precision than any human ever could achieve before, which has led to still more innovation. We can do things like search the Internet instantaneously for information stored on the other side of the planet, and lob exploration robots at Mars and monitor the data they send back. The presence of more and more functional robots could lead to innovations that aren't even occurring to us right now.

Potential Future of Industrial Robotics

Toyota unveiled a violin-playing robot in 2007 in Tokyo, but artists probably don’t need to fear for their jobs. The performance was designed to showcase the precision of movement the robot is capable of.
Toyota unveiled a violin-playing robot in 2007 in Tokyo, but artists probably don’t need to fear for their jobs. The performance was designed to showcase the precision of movement the robot is capable of.
©Koichi Kamoshida/Getty Images

In the 1980s there were only a few thousand industrial robots in operation in the U.S. [source: Condon]. Now there are hundreds of thousands, and more than a million worldwide. Many of them are working in the automotive industry, moving, cutting, drilling, welding, painting and otherwise assembling our cars, but they are also employed in pretty much every other industry. With its ease of use and low price point, Baxter might do for industrial robotics what Roomba did for in-home robotics and increase the numbers still further.

Baxter is being aimed at the nearly 270,000 small to midsize manufacturers, which have five hundred or fewer employees [source: Freedman (Inc)]. Companies of that size are unlikely to be able to invest hundreds of thousands of dollars into robots that require a redesign of their workspace and IT personnel to run them, but Baxtermight be within their reach. And with such companies, the robot's initial reception has been good. Nypro, an injection-molding firm, and Vanguard Plastics have test-driven Baxter and expressed interest in putting it to work in their plants. And larger companies could also consider it for cost-effectively automating tasks they are currently doing manually.

Baxter isn't the only game in town, although he's likely the least expensive. There are a few other robotics companies trying to get into similar markets. The Nextage robot by Kawada Industries is a two-armed humanoid assembly robot that can work alongside humans like Baxter. It's more precise, but costs nearly $100,000 [source: Toto]. Barrett Technology has a compliant arm with seven degrees of freedom, and back-drivable and force-sensing capabilities, also for around $100,000. Universal Robot has human-safe and easily programmable robot arms for around 20,000 Euros (roughly $27,000) that can do heavier lifting than Baxter. Yaskawa Electric Corp's dual-armed Motoman robot has been demonstrated dealing blackjack. And Fanuc Corp and ABB Ltd, makers of heavy industrial robots, are reportedly working on smaller models [source: Guizzo and Ackerman]. So, there will be lots of choices to get more robots into businesses working right alongside humans in the near future.

Provided we can work out the socio-economic issues so that we all stay gainfully employed, this new technological revolution will allow us to come up with new and creative things for ourselves and our robot co-workers to do. If robots can take over more undesirable or difficult tasks, the sky's the limit for human achievement. Many of us can shift away from mindless production to knowledge work. Robots are even making forays into non-industrial areas like assisting with surgery, X-ray scanning, bomb detection and search and-rescue work. As their uses increase, so will demand. Perhaps there are future jobs to be had by us all in the field of robotics itself.

Author's Note: How Baxter the Robot Works

Robots are cool and I'll never tire of reading about them. I may even make one someday. I have an old LEGO Mindstorms kit that's been gathering dust, and a couple of new Raspberry Pi computers. I just need to set aside some time and drain my bank account on a few more items (like robotics books) before getting started.

One jarring revelation from my research, however, was that a software application that writes articles exists. It hadn't occurred to me that we were already at the point when a program could generate a readable story. If it progresses beyond sports and stock write-ups, I may soon know what it feels like to be automated out of a job.

My other vocation, computer programming, may not be safe, either, if the user-friendly programmability of robots like Baxter catches on in other areas. Maybe fiction writing, my other-other vocation, will be harder for an artificial intelligence to emulate. But it's also more difficult to turn into a paying gig. We may need to switch from capitalism to some form of utopia for this all to work out.

Still, I'm all for technological progress, and especially for the removal of drudgery from work via robotics. There are much better ways we can spend our time. Like exercising to prevent our newfound lack of activity from turning us into doughy layabouts, or studying up for those impending career changes.

Related Articles


  • Athawale, Vijay Manikrao and Shankar Chakraborty. "A comparative study on the ranking performance of some multi-criteria decision-making methods for industrial robot selection." International Journal of Industrial Engineering Computations. November 2011, Volume 2, Issue 4, Pages 831-850. (January 6, 2013)
  • Atkinson, Robert D. "Building a More-Humane Economy." Futurist. May/June 2006, Volume 40, Issue 3, Pages 44-49. (January 6, 2013)
  • Barrett Technology, Inc. "The New WAM Arm." (January 15, 2013)
  • BBC News. "Robotic co-worker Baxter joins factory line." September 18, 2012. (December 31, 2012)
  • Brooks, Rodney and Anita M. Flynn. "Fast, Cheap and Out of Control: A Robot Invasion of the Solar System." Journal of the British Interplanetary Society. 1989, Volume 42, Pages 478-485. (January 15, 2013)
  • Condon, Mary. "Straight Talk About Robots." Training & Development Journal. November 1983, Volume 37, Issue 11, Pages 14-24. (January 6, 2013)
  • Deutsch, Steven. "Automation: A Study of Local Union Leaders." Bulletin of the Atomic Scientists. May 1968, Volume 24, Issue 5, Pages 52-53. (January 6, 2013)
  • Dibbell, Julian. "The race to build intelligent machines." Time. March 25, 1996, Volume 147, Issue 13, Pages 56-58.
  • Economist. "Baxter gets to work." September 29, 2012. (January 3, 2013)
  • Flick, Frank. "Does 'Automation' Cause Unemployment?: EFFECTS OF AUTOMATION." Vital Speeches of the Day. February 15, 1962, Volume 28, Issue 9, Page 280-281. (January 6, 2013)
  • Frank, Aaron. "Could Automation Lead to Chronic Unemployment? Andrew McAfee Sounds the Alarm." Forbes. July 19, 2012. (January 10, 2013)
  • Freedman, David H. "Bringing Up RoboBaby." Wired. December 1994, Issue 2.12. (January 13, 2013)
  • Freedman, David H. "The Rise of the Robotic Work Force." Inc. October 2012, Volume 34, Issue 8, Pages 76-83. (December 31, 2012)
  • Gillis, Charlie. "The robot working-class invasion." Maclean's. October 17, 2012. (December 31, 2012)
  • Guizzo, Erico and Evan Ackerman. "How Rethink Robotics Built Its New Baxter Robot Worker." IEEE Spectrum. October 2012. (January 1, 2013)
  • Guizzo, Erico. "World Robot Population Reaches 8.6 Million." IEEE Spectrum. April 14, 2010. (January 15, 2013)
  • Hagerty, James R. "Baxter Robot Heads to Work." Wall Street Journal. September 18, 2012. (December 31, 2012)
  • iRobot. "About iRobot." (January 14, 2013)
  • iRobot. "Defense and Public Safety." (January 12, 2013)
  • iRobot. "Your Home, Our Robots." (January 12, 2013)
  • Jackson, Stanley. "Rethinking Humanoid Robotics." Manufacturing Digital. December 18, 2012. (January 12, 2013)
  • Kelly, Kevin. "Better Than Human: Why Robots Will - And Must - Take Our Jobs." Wired. December 24, 2012. (January 12, 2013)
  • Kuzma, Cindy. "The Human Touch." Science and Spirit. July/August 2006, Volume 17, Issue 4, Pages 15-17. (J, 2
  • Levy, Steven. "Can an Algorithm Write a Better News Story Than a Human Reporter?" Wired. April 24, 2012. (January 12, 2013)
  • Lorincz, Jim. "Technologies Offer Process Variety: Waterjet, PKM and robots with red faces offer many choices." Manufacturing Engineering. November 2012. (January 12, 2013)
  • Markoff, John. "A Robot With a Reassuring Touch." New York Times. September 18, 2012. (January 14, 2013)
  • MIT Artificial Intelligence. "Cog - Overview." (January 13, 2013)
  • MIT. "Rodney Brooks - Roboticist." (January 14, 2013)
  • Mraz, Stephen J. "A robot for the rest of us." Machine Design. October 16, 2012. (December 31, 2012)
  • Murphy, Robin, Red Whittaker, Javier Movelian, Rodney Brooks, and Corey S. Powell. "Machine Dreams." Discover. May 2010, Volume 31, Issue 4, Pages 30-76, 6 Pages. (January 13, 2013)
  • Nelson, Rick. "Robots, jobs, and war." EDN. September 17, 2009, Volume 54, Issue 18, Page 6. (January 6, 2013)
  • New Scientist. "Designed for life." June 1, 2002, Volume 174, Issue 2345, Page 46. (January 13, 2013)
  • New Scientist. "Robot workmate." September 22, 2012, Volume 215, Issue 2883, Page 4. (December 31, 2012)
  • OpenCV. "About." (January 15, 2013)
  • Rader, Louis T. "Automation Over the Years." Vital Speeches of the Day. January 1, 1981, Issue 6, Pages 166-171. (January 6, 2013)
  • Raskin, A.H. "The Great Society - Its Impact of Automation" Vital Speeches of the Day. July 1, 1966, Volume 32, Issue 18, Pages 554-559. (January 6, 2013)
  • Rethink Robotics. (January 12, 2013)
  • Rethink Robotics. "Baxter." (January 4, 2013)
  • Rethink Robotics. "how baxter is different." (January 12, 2013)
  • Rethink Robotics. "Rethink Corporate Brochure." (January 4, 2013)
  • Rethink Robotics. "Solutions." (January 12, 2013)
  • ROS. "Documentation." (January 15, 2013)
  • ROS. "Robots Using ROS." (January 15, 2013)
  • Smock, Doug. "Nypro hopes humanoid robots kickstart reshoring." Plastics Today. September 24, 2012. (January 12, 2013)
  • Toto, Serkan. "Japanese Company Shows Robot Co-Working With Humans (Video)." November 28, 2011. (January 15, 2013)
  • Universal Robots. "Universal Robots' products." (January 15, 2013)