According to Aaron Prather of ASTM International, the IEEE group’s chairman and director at ASTM International, one distinguishing feature is that humanoids are “dynamically stabile”. They need power to remain upright, and they use their legs (or other limbs), to exert force to stay balanced. Prather explains that in traditional robotics if anything happens, you simply hit the little button to stop the power. “You can’t do that with a robot. If you do it, the robot is likely to fall and pose a greater risk.”

Slightly slower brakes

How might a safety feature appear if it is not an emergency stop. Agility Robotics has added some new features to the latest version Digit in an attempt to solve the toppling problem. Instead of instantly depowering and likely falling down, the robot could decelerate gently when, for example, a human gets too close. Velagapudi explains that the robot has a limited amount of time in which to reach a safe condition. It could put down any items it is carrying and get down on its hands and knees, before shutting down.

Different robotic systems could approach the problem in different manners. Federico Vicentini is the head of product safety for Boston Dynamics. He says, “We want standardize the goal and not the method to achieve it.” Vicentini chairs a working group of the International Organization for Standardization to develop a standard for the safety and stability of industrial robots. Experts from Agility Robotics also participate. He says the idea is to establish clear safety expectations while not limiting innovation by robot and component manufacturers. “How to solve this problem is up to you, the designer.” How do you define a humanoid robotic system? Does it have to have legs? Arms? A head?

Prather: “One of our suggestions is that we might need to drop the term ‘humanoid’ altogether.” His group proposes a classification system that would consider the capabilities, behavior and intended uses of humanoid robotics, rather than their appearance. The ISO standard Vicentini has been working on refers all industrial mobile robots that have “actively controlled stability.” This could apply to Boston Dynamics’ doglike quadruped spot as well as its bipedal humanoid Atlas and could also cover robots with wheels.

Learn to speak robot

Humanoids are not only a safety concern, but also a communication problem. They will have to communicate their intentions to everyone in a way that is understandable if they are to share the space with humans. Just as cars use brake and indicator lights to show their intention, so too will they need to be able to recognize when someone is about to cross over their path. Velagapudi says Digit has lights that show its direction and status, but better indicators are needed if it is to work collaboratively with humans.

If Digit is going to walk into an aisle right in front of you, then you don’t want that to surprise you. The robot can use voice commands but audio alone would not be practical in a noisy industrial setting. It could even be more confusing if there are multiple robots in a space. Which one is trying to grab your attention? Prather says that there is also a psychological difference between humanoids and other types of robots. We tend to anthropomorphize robots who look like us. This can lead us overestimate their capabilities and become frustrated if they fail to meet our expectations. “Sometimes, you let down your guard on safety or your expectations about what the robot can do versus the reality,” he says. These issues are particularly problematic when robots will be performing emotional labor or providing support to vulnerable people. The IEEE report recommends any standards include emotional safety assessments and policy that “mitigates psychological stress or alienation.” People wanted robots with facial expressions and the ability to read micro-expressions. They also wanted robots who could communicate using gestures, voice and haptics. “They wanted everything–something that doesn’t exist,” she says.

Escape the warehouse

Human-robot interaction could be crucial if humanoids want to move from industrial spaces into other contexts such as hospitals, elderly living environments, or homes. Hilburn says that robots who may be working with vulnerable groups are of particular importance. “The damage that could be done in an interaction with a robotic if it is not programmed to talk in a manner to make a person feel safe, whether it is a child or older adult, can certainly have different outcomes,” she says. The IEEE group has made recommendations that include enabling human override and standardizing visual and auditory clues. They also recommend aligning the robot’s appearance to its capabilities in order to avoid misleading users. Prather says that if a robot appears human, people will expect to be able hold a conversation with it and show some emotional intelligence. If it can only perform basic mechanical tasks, it could cause confusion, frustration and a loss in trust. He says, “It’s like self-checkout machines.” “No one expects these machines to chat with you, or help you with your groceries. They’re clearly machines.” If they appeared to be a friendly employee, but then repeated “Please scan your next item,” people would become annoyed.

Prather & Hilburn both emphasize that inclusivity and adaptability are key when it comes human-robot interactions. Can a robot interact with blind or deaf people? Will it be able adapt to people who need more time to reply? Can it understand accents of different languages? Prather says that robots operating in different environments may need to adhere to different standards. A robot that works in a factory with people who are trained to interact is one thing. But a robot that is designed to assist in the home, or to interact with children in a theme park, is quite another. The public will be able to better understand robots if they have some ground rules. He says it’s not about being prescriptive, but rather about establishing some basic guidelines, so that manufacturers, regulators and end users know what to expect. “We’re simply saying you’ve gotta hit this minimum bar – and we all agree that below that is bad.” Vicentini says it’s too early to talk about humanoid robotics. We haven’t seen their current state of development, but it’s important to put some checks and measures in place. Standards help manufacturers gain trust in their products, making it easier to sell in international markets. Regulators often rely on standards when establishing their own rules. Vicentini says that it is difficult to create a standard that everyone can agree on because of the diversity of players. “But everybody being equally unhappy is enough,” he says.


www.roboticsobserver.com