February 09, 2017      

Right now, 6 million people in the U.S. alone are working in retailer warehouses, and 50 percent of their time is spent shifting products. Human workers spend 25 percent of their time taking empty carts back to where they were. To improve productivity and engagement, supply chain companies are increasingly turning to automation and mobile robots, but good human-robot interaction design is essential.

In my previous articles, we looked at how Willow Garage led research into human-robot interaction (HRI) and how the robot design principles learned there have guided makers of commercial service robots.

Business Takeaways:

  • Interaction design isn’t just for making service and social robots attractive; it’s needed to optimize robot functions and adoption.
  • Examples of commercially successful robot designs incorporated the lessons from research at Willow Garage into functionality and human-robot interaction.
  • Before robots are accepted alongside humans in daily life or fade into the background, robot makers need to understand the user needs they must address and how their roles affect their design.

Unlike the examples of Savioke’s Relay or Simbe Robotics’ Tally, the mobile robots from Fetch Robotics are intended for use inside warehouses rather than dealing with the public. Because of that, its interaction design is as simple as possible.

You can choose commands such as “Follow me,” “Go to charging station,” or “Stop” from the tablet interface, and Fetch and Freight are programmed to stop working if something unexpected happens.

The mobile picker and conveyor move simply. When Freight is told to follow, it looks at the worker’s footsteps and follows at an appropriate distance. It can go around obstacles without bumping into them.

Following robot design principles, Fetch and Freight are non-threatening.

Fetch and Freight have different interaction design than customer-facing robots.

If the robot needs to follow a group of people, it simply follows the nearest person, regardless of who gave the original command. There is little close interaction among warehouse personnel, and the chance of someone trying to steal the robot is lower than in customer service.

E-commerce order-fulfilment staffers need to register the map of the space in which the robot is allowed to move. The idea is that they can choose where they don’t need the robot to go, reducing unnecessary data and processor cycles.

Freight is designed to stop operating and goes back to its power station if the actual environment is different from the previously registered map. This is to prevent the robot from trying to figure out a dynamic environment that it’s not prepared for and going out of control.

As the Amazon Picking Challenge has demonstrated, autonomous robots have difficulty distinguishing between the objects that are very similar, such as a Coke can and a Pepsi can.

Instead of making its robot more clever than necessary, Fetch Robotics encourages users to adjust their settings depending on the role of the robot. The company recommends grouping similar objects by location in the warehouse. This recommendation to have robots make as few decisions as possible is based on the idea that robots are not perfect.

From B2B to B2B2C, needing good interaction design

In comparison, Savioke appears to have spent the most engineering resources on interaction design. That is because Relay is the only the robot that supposes to do services to customers directly.

Unlike robots in factories, which are often controlled by engineers, the users for logistics robots are typically non-engineers. Moreover, the service robots coming to our daily lives will be operating alongside the unskilled general public — including children.

Robotics engineers need to adapt the level of interaction design based on the environment and people surrounding their devices.

One example of such an adaptation is Anki Inc., a business-to-consumer (B2C) company that adopted the animation development method when it developed the interactive toy robot Cozmo.

Willow Garage’s legacy in interaction design

As with the development of applications for the Web and mobile devices, robot design is still evolving. For robotics to be accepted into our daily lives, designers must gather information about how people will react to and use them.

As we’ve seen, the interaction design principles studied at Willow Garage have been applied in robots from startups Fetch Robotics, Savioke, and Simbe Robotics. The following common rules for specialized service robots can be seen within these designers’ work:

  • Make robots’ physical form obvious what the robots can do and what they are doing from their physical form.
  • Keep humanlike capabilities of the robots to a minimum to avoid unrealistically high expectations.
  • Create opportunities for people to interact with robots in the environment using expressions and sound effects.
  • Make the appearance friendly, not to give any threatening or frightening expressions.
  • Make sure the robots follow the rules in human society, such as facing forward in an elevator.
  • All functions and communications need to be as clear as possible.

However, the different settings for human-machine interaction — such as retail stores, warehouses, hotels, and shopping malls — and the unique tasks performed by robots necessitates unique commercial products.

The former Willow Garage designers acknowledged that they made mistakes early on and that their robots have only reached the market after extensive they prototyping and user tests. Even then, they might say they are still beta version.

When I asked these robotics leaders, “What is the secret of interaction design in the era of service robots?” they all gave the same advice: “Talk to your customers. Let them try them and see how they react. Robots never come first. It is always ‘customers first.'”

More on Service Robot Design:

A scenario for the robots to adapt themselves

If you compare the design of service robots such as SoftBank’s Pepper or Sharp’s RoBoHoN, a robot whose exterior doubles as a smartphone and follows a “Robot first” concept, you may find operationally specialized, nonhumanoid robots such as Relay, Tally, or Fetch to be unattractive.

It’s true that the exteriors of consumer-facing robots are designed to be attractive, and it’s easy to mistake that for the key to adoption. On the other hand, when some products and services become fully accepted in society, they may no longer be acknowledged and gradually disappear.

Human telephone operators are one example. Their role has been automated, and people no longer ask to be connected to someone; they just dial a personal number or ask their smartphones to dial. Could some robots eventually be taken for granted or fade into the infrastructure?

Contrary to a constant media drumbeat that robots and artificial intelligence threaten humanity’s future, the main challenge facing social and service robot makers is to understand what people need from robots. Beyond creating nonthreatening robots, those who follow a “customer first” strategy for interaction design will be able to improve robot function, acceptance, and integration into a company and society.