Robotics Moves To The Mainstream

Don’t look now, but an autonomous robot might be around the corner cleaning the floor, delivering medical supplies, or performing some other useful job. You literally could bump into a robot while it’s trying to complete its chores for its human overlords. You might even be one of those overlords.

Robots have been improving for decades, but what started in a university lab is now available at the hardware store, like iRobot’s Roomba floor cleaner (Fig. 1). If you’ve ever visited the University of Maryland Medical Center (UMMC), you may have seen the Aethon TUG delivering everything from medicines to meals (Fig. 2).

These robots are doing monotonous jobs like vacuuming that people used to do. They could be seen as interlopers taking jobs away from people, or they could be seen as tools that augment what people can do.

Robots are used in assembly lines to create everything from cars to cookies, but they’re typically isolated from people or put inside a cage for safety. It’s not uncommon to see a robot moving at a blur behind a plexiglass window. Sensing systems such as light curtains make environments that are more open possible, shutting down the robot if an object or a person crosses the curtain.

Sometimes robots are segregated because the work environments are toxic or dangerous. Sometimes, they’re segregated because they can’t identify people and could hurt them. However, some robots have to interact with people, making safety solutions much more difficult. Hardware and software improvements are helping with increased flexibility. 

Technical improvements such as improved sensors (see Safe Robots Rely On Sensors) and greater computing power also have influenced the design of robots that can work with and around people. Robots now can determine when people and other objects are nearby and how they should react. The level of sophistication is still limited in most cases but usually sufficient for the target applications. 

Changing The Assembly Line

Robots on the assembly line normally have been assigned simple, repetitive tasks that they can perform better than most people. Falling prices and improved performance make these robots more cost-effective in some areas. But for complex products like cars, robots complement people who handle other tasks along the assembly line. A large robotic arm can easily lift a heavy car hood and place it on a car moving down an assembly line where it may have taken two people to do the same job.

On the other hand, handling more intricate wiring or placing an object in a difficult to reach spot may be beyond a robot’s capability. Robots and people have usually been separated along these kinds of assembly lines. Likewise, the robots are normally programmed to do one task, and that programming tends to be precise.

Rethink Robotics is targeting its Baxter robot at assembly environments where interaction with people is more common, including programming and operation (Fig. 3). It could sit in the corner and churn away forever, but it’s more likely to be tuned or even re-tasked as necessary. It’s also designed so whoever is doing the re-tasking doesn’t need a degree in programming or extensive technical training to handle most jobs. Baxter is designed to be human-friendly, too.

“Baxter is a totally new kind of robot. It is a robot with common sense, so it knows what you want and does what you expect,” says Mitch Rosenberg, vice president and product management at Rethink Robotics. “Baxter is aware of its surroundings and can adapt as needed. One particularly important example of its common sense is its awareness of and adaptation to humans in its environment. This makes it safe to work alongside humans.”

Baxter’s counter-weighted arm design allows low-impact operation. When Baxter accidentally hits you, it hurts about as much as bumping into another person. If other robots typically found on an assembly line hit you, though, you’re likely to get a fracture.

Baxter uses behavior-based programming, just like the Roomba, but Roomba works on a much simpler scale. Behaviors are rules that consider the current state of affairs including inputs from sensors as well as computer-based and user-based directives. If a rule is met, then an associated action is performed. Rules may be simple, such as “if a front sensor indicates an object then stop forward motion.” Rules can be combined to perform complex functions and provide sophisticated behaviors from an external point of view.

It’s possible to get down to this level of programming, but the system is designed to use the behaviors to implement a high-level programming environment including an interactive training system. In this case, it can be as simple as pressing the program button on an arm, positioning the arm by holding the hand and moving it, pressing the program button again to let the hand grab an object, lifting the arm and moving to a new location, and pressing the program button again. The arm can then repeat this process of picking up an object and moving it to a new location forever.

Rethink Robotics provides a software development kit (SDK) for more complex or custom programming chores. The SDK can be used to augment the interactive programming system and provide ways for the robot to interact with other devices within its environment.

Pick-and-place applications are common in assembly line jobs, though these applications don’t take advantage of Baxter’s other capabilities. For example, the hand can have a camera that can recognize the shape and orientation of an object so it doesn’t have to be found in a fixed location, which pick-and-place robots require. In more complex applications, Baxter could perform different actions depending on the type, color, and orientation of an object. It could sort red objects into one bin and green into another, for instance.

Baxter’s behaviors also address safety issues such as collisions between its two arms. If one arm is programmed to move objects from point A to point B and the other is programmed to move objects from point B to point C, then Baxter knows not to have both arms at point B at the same time. It’s also impossible to move the two arms so they touch when you’re programming Baxter.

Baxter isn’t designed for heavy loads. Its typical hand payload is five pounds. It moves about 2 feet/s under load and 3.3 feet/s without. The arms have seven degrees of freedom, which is on par with a human being’s range of motion. 

Its hands, also called end-effectors, are a bit more flexible than a human hand. They accept easily interchangeable devices from clamps to vacuum cups. The hands and arms have force feedback for safety and more precise operation.

Additional cameras, sonar sensors, and a flat panel display are located where one might associate a head. The cameras and sonar can identify objects around the robot, whereas the hand cameras are used for identifying items below or in front of the hand.

Baxter is intended for assembly chores that might change on a regular basis, such as short production runs or handling different objects in a specific manner. These changes would be impractical with a conventional assembly line robotic system, but Baxter excels in these environments.

Baxter is an articulated robot but not mobile. It is designed to sit in front of an assembly area. Mobile robots with arms are commonly used in military applications and even police and rescue applications, but they tend to be semi-autonomous teleoperated robots like unmanned aerial vehicles (UAVs) and explosive device removal robots. At this point, the robots that you might encounter couldn’t give you an actual hand.

Robots Help Hospitals

Aethon has quite a few TUGs that have found work in hospitals. They shuttle carts throughout the hospital, picking up and delivering everything from linens to food and medicine. The TUGs and carts tend to be optimized to specific tasks, but their operation and interaction with people is identical.

The TUGs have a different challenge than Baxter but many of the same concerns. Safety and flexibility are at the top of the list, and improved sensors and computing power have made safe and efficient operation possible.

The TUG is an electric robot that can pull a cart. Its four sealed lead-acid batteries support a pair of electric motors for about 10 hours. The drive wheels have a patented ultra-accurate odometer system. The system has a 31.7-in. turning radius and a top speed of 225 feet/minute. And, the TUG can handle a 10% grade and a load up to 1000 pounds.

The robot has overlapping laser, sonar, and infrared sensors. Wi-Fi provides communication with the central management system, which tracks the location of all the robots. The wireless link is also used to control other devices such as doors and elevators.

The TUG operates in the hospital using the same doors and corridors as people. It is polite and makes way for people and other robots. The robot uses its own elevators in some hospitals, but more for efficiency, not because it needs to be isolated. Charging systems are normally deployed where TUGs can congregate when they aren’t active.

TUGs move their cargo as necessary, such as picking up meals at the cafeteria and then stopping at the appropriate rooms or stations, where a person will remove the meals for final delivery to a patient. TUGs also are used to return the dirty dishes. The same operation occurs for linens and medicine, with different destinations.

Of course, the carts are a bit different. Medicines require a locking system that would be programmed so they are dispensed to the correct patients. Likewise, people interacting with a TUG at a designated location will need to indicate if anything was removed or added to the cart because the TUG might not have sensors that could register these changes. The UMMC deployment uses the MedEx tracking system and radio-frequency identification (RFID) plus biometric scans to create an electronic chain of custody. This approach reduces paperwork, increases security, and complies with medical practices and laws.

The system employs a minimum user interface normally limited to large buttons and some audio visual feedback. It’s easy for any of the hospital staff to work with, but patients and visitors wouldn’t use it. The hospital management system keeps track of all the robots and their loads so personnel know when and where they will be. For example, nurses could receive a text indicating a robot will be delivering lunch in a couple of minutes.

Unlike Baxter, Aethon is involved in site design. The TUGs are versatile but they do have some operational requirements. Most hospitals are already designed to meet these requirements, though. For example, electric doors are common throughout hospitals. The system also must fit the hospital’s existing methodology. This means studying the current flow and determining how the TUGs would integrate with that environment. Even simple optimizations can result in significant savings.

Aethon’s system can work in other environments, but its current thrust is in the medical space where the company has developed a high level of expertise. In the first year at UMMC, the system reduced cycle time for deliveries from 74 minutes to 30 minutes. The robots are available 24/7 so even the time of day doesn’t affect them. Hospital personnel are more efficient since they don’t have to track the deliveries the TUGs make, allowing them to spend more time with patients.

Robots Underfoot

The Aethon TUGs interact with humans by navigating around them as well as through their button and display interface. There are usually lots of TUGs in those environments as well. On a more personal level, people can buy robots like iRobot’s Roomba to take care of their floors.

Robots designed for home use are relatively simple to operate, with a couple of buttons and LEDs for an interface. They often use a random pattern to cover an area, and they’re so small you need to avoid stepping on them. They typically recognize obstacles and steps but little else. They beep when they’re done, and they can dock themselves for recharging.

If you want to experiment with a domestic robot, check out the iRobot Create. A computer including something as small as a Gumstix module can control it (see “TurtleCore Tacks Cortex-A8 On To iRobot Create”at electronicdesign.com). The vacuum support has been removed, though, to make room for your hardware.

The same principle applies in slightly larger packages for mowing lawns, with products available from BelRobotics, LawnBotts, and Robomow (Fig. 4). The batteries and frames in these electric mowers are a bit larger than the vacuum cleaning robots. They also can do a bit more damage to people, so they have a more robust sensor system that probably makes them safer to use than conventional lawn mowers, which depend on people for control.

The safety system for these mowers is impressive. The blades are designed to utilize less power, increasing their run time and the area they can cut. Likewise, the designs account for varying grass levels. The large robots like the BelRobotics BigMow can handle as much as 5 acres.

Robots Take Your Place

You’re also likely to run into telepresence robots like Bossa Nova’s mObi (Fig. 5). These robots can move in and around people. They also typically have a flat-panel display where the remote operator can be seen and a camera to show the operator what the robot sees. Audio support completes the design.

The robots tend to be about four or five feet tall, bringing the camera and display near an adult’s eye level. They have an array of sensors to detect people and other objects so they can move about without colliding with them.

From a safety point of view, these robots tend to be on par with the vacuum cleaning robots. They are small and light enough not to pose a major threat to people, although they could be a problem around small children. Usually the collision avoidance is designed to keep the robot from getting hurt.

These robots tend to look different, but the main differences really are in the drive system. Bossa Nova’s mObi rides on top of a ball, taking the balancing act of the Segway Human Transporter to the next level. The inverted pendulum is simply on a ball and sensors, and the drive motors are placed accordingly.

Like the Segway, the mObi requires very little power to remain balanced. It also tends to be efficient in moving. It has the advantage in omnidirectional movement and recovery that other telepresence robots lack. Bump into the mObi, and it will move. Bump into another telepresence robot, and it might not.

The Bossa Nova mObi has a lot of competition from Anybot’s QB (see “Any Bot In A Telepresence Storm”at electronicdesign.com) to the VGo (see “Cooperation Leads To Smarter Robots”at electronicdesign.com). Suitable Tech’s Beam is going to be used to remotely direct a motion picture (see “Interview With The Producer Of Tower Of The Dragon”at electronicdesign.com). The telepresence robot provides a mobile viewing platform, but the digital recording of the movie enhances the telepresence approach since it delivers an additional view.

Telepresence robots are teleoperated by definition, but they’re normally semi-autonomous. Operators often can take over full control, but they’re more likely to specify waypoints with the robot doing the rest. This includes path planning and collision avoidance.

Robots aren’t everywhere yet, but the technology is starting to show up in all sorts of places including cars. We tend to count the number of electronic devices we have, from smart phones to HDTVs. In the future, we might count the number of robots we have.