Humanoids are putting India on the global robotics map. Desi humanoid robots are turning up at banks, offices and events. They’re still simple creations, but they show the way forward, say their makers. The eventual aim? To be bot suppliers to the world
If you’re on Beach Road in Puducherry, be sure to say hi to Constable Singham, a smiling, stationary humanoid with a statement moustache and crisp khaki uniform. He is programmed to tell you about the nearest tourist attractions, or direct you to the nearest hospital.
At branches of HDFC, Canara and City Union banks, customers can expect to be greeted by a humanoid who can tell them how to open an account, or guide then to the right counter.
If all goes well, in 2020, the sales lounges of Puravankara builders in Bengaluru, Chennai and Pune will also have humanoids instead of receptionists, answering basic questions about their projects.
Humanoids, or robots that look, talk and sometimes move about like human beings, are making their first forays into public spaces in India. It’s an important step, say researchers, because it’s putting India on the global robotics map.
“This first wave of humanoids is a platform for research. It tells us what’s working and what’s not,” says Arun Giriyapur, head of automation and robotics at the KLE Technological University in Karnataka.
And that’s crucial because humanoids could be the next thing the world out-sources to India — basic low-cost bots that bring you your spectacles, take your blood pressure at the doctor’s office — made in India and exported to countries that need them.
The aim is full-time deployment in areas such as customer service, security patrolling and healthcare.
In Japan, with its rapidly aging population, humanoids are already relatively popular as helper and companion robots for seniors.
In education, they can help with language-learning, and overall learning for children with developmental disorders such as autism, since they can react in a consistent manner repeatedly and patiently.
Our bots are still learning. KLE University’s Ajit 1.0, developed last year, is 2-feet-tall and can shake hands and do a namaste in response to voice requests. A five-year project is aimed at improving his appearance and skills.
Indro, three versions of which have been built since 2016 by Santosh Hulawale, an independent innovator, is 7-feet- tall, and can perform lightweight household tasks such as bringing you things that weigh less than 2 kilograms — but only if they are in the exact location where he has been programmed to find them. Research is on to make him better at finding things.
Five-feet-tall Mitra (female version is called Mitri), created by Bengaluru-based Invento Robotics, is one of the few humanoids with a full-time job. He’s now Interactive Robotic Assistant (IRA) 2.0, a receptionist at the Koramangala branch of HDFC Bank in Bengaluru. Kerala-based Asimov Robotics’ 5.5-feet-tall Sayabot was their IRA 1.0.
These robots can navigate autonomously, understand voice input, recognise faces and use Artificial Intelligence (AI) for comprehension, sort of like the voice-based operating systems Alexa and Siri.
The IRAs can live-tweet, greet guests and tell customers about a product in multiple languages. But they cannot carry heavy objects, navigate stairs, or answer subjective, complex questions — like “how do you like your job?”
When you think of a humanoid, says Jayakrishnan T, the founder and CEO of Asimov Robotics, “people tend to think of what they have read in books or watched on screen”. And so there are questions about potential threats — will robots take all the jobs? Enslave humans? Run the world?
The reality is that most robots around the world have the comprehension of a four-year-old and mobility levels of a six-year-old, says Balaji Viswanathan, CEO of the Invento Robotics. “Where they can do more, as in the case of Sophia, they are not commercially feasible.”
Sophia was revealed in 2016 by Hong Kong-based Hanson Robotics. She waves, jokes, winks and smiles (but can’t walk). Meant to look like Audrey Hepburn, she quickly became a celebrity herself and has ‘chatted’ at press interviews and appeared on The Tonight Show with Jimmy Fallon.
Sophia is the only one of her kind and costs about Rs36 lakh per event to rent. And even she is not a true form of AI. She uses the same technology that most chatbots use — drawing from a set of a pre-programmed responses. Admittedly, her set is massive, and far more complex. But she cannot retain new information or use it. For instance, she may ask your name (because she’s been programmed to), but she can’t ‘remember’ it or use it in conversation.
“In India, where manual labour is abundant and sadly available at extremely low rates, there is even less of an incentive to invest millions in replacing people with machines,” says Hulawale, inventor of Indro.
Asimov Robotics, for instance, took three years and over ₹ 35 lakh to create the first prototype of their humanoid Sayabot, from scratch.
“The first one always costs more. I could make the subsequent model for about half of the cost,” says Jaykrishnan T. That would still mean Rs15 lakh-25 lakh per bot.
“Today, customers are dabbling with various technologies and platforms and what is keeping them hooked is the experience they derive from them. Banking is no different. We are using technology to enhance customer experience and IRA 1.0 was a step in that direction,” says Nitin Chugh, country head – digital banking, HDFC Bank. “We currently have two humanoids; one in Kochi and the other in Bengaluru. We will continue to assess the potential of technologies, where they complement the efforts of the branch staff.”
Most of the current global demand is in the retail space, followed by offices (mainly as receptionists), companion robots and the education sector.
‘Can i help you?’
Customer service robots can attract a lot of attention, says Doguparthi Lekha, founder of a Hyderabad-based health-food startup named DietLabs. “At the launch of our first stall in Hyderabad earlier this year, we hired Mitra to help draw in the crowds. It worked! Admittedly, there were more questions about him than my product,” she says, laughing. “And there were glitches. Mitra couldn’t understand some accents. His speech was a bit slow. But even today, we’re known as “the ones who hired a robot, no?’.”
For Anand Narayanan, COO of the Puravankara Group, which used Mitra at a property expo in July, it was novelty and efficiency that made it worth the cost. “Customers really responded to it,” he says. “It also helped streamline the queries.”
Traditionally, a few sales people greet customers, answer their property- related questions and take down these details at an event. “This information is then fed into a system, after which our customer relationship managers make follow-up calls, about 100 hours later,” Narayanan says.
“Sometimes the handwriting is not legible; the energy levels of the sales people change depending on the time of day. The humanoid, on the other hand, was consistent, and data was transferred in real time.”
“The novelty of the current batch of humanoids will wear off,” says Jaykrishnan of Asimov. “As scientists, we are excited about exploring how far we can go in creating the perfect humanoid — that mimics a human, shows emotion and understands context in conversation.”
The ‘perfect’ humanoid should have artificial intelligence that can independently drive speech and comprehension. Interestingly, it should be human-like but not human-identical — studies have shown that interacting with a machine that looks and sounds somewhat human raises engagement levels; but a bot that looks too human, they found, causes discomfort and unease.
Even with the right AI, the real challenge, say manufacturers, is creating a humanoid that can move, speak, understand — and multitask.
The most expensive humanoids in the world can’t do more than one thing well. A chess-playing robot can’t answer questions. A customer service bot can listen and respond but can’t begin to kick a ball or strategise for game-play.
“We’re looking at creating bots that can search the Internet to fill in gaps in their own knowledge, but even that has serious limitations,” says Jaykrishnan. “You might ask a bot for your glasses and be handed a tray of drinking water instead.”
So why bother?
According to government estimates, 35 per cent of India’s population will be between 18 and 35 by 2020. “Because of our young population, there is no reason to invest in teaching something to a bot when a human could do it better,” says Milan Sheth, advisory partner and leader of intelligent automation at Ernst & Young.
There is a point in being the country that makes good humanoids for the world. “India has always had software talent, and robotics is no exception,” says Giriyapur. “It’s the hardware we need to work on.”
For instance, both Mitra and Sayabot, which can be considered India’s counterparts to Japan’s Pepper, currently the most widely used customer service robot, cannot move their heads as Pepper does.
“The Japanese started humanoid research several decades before the rest of the world, have made much larger investments and have a history of technological superiority,” says Viswanathan of Invento.
It’s a cycle. Bigger investments, due to economies of scale, will make humanoids cheaper, and create the demand that will in turn push research and more production. “India doesn’t have the kind of demand that would create such an ecosystem,” says Kisshhan Psv, founder of H-Bots. But demand is already coming in from elsewhere.
Hyderabad-based H-Bots has built a Robocop complete with painted-on uniform, cap and gloves. It will be starting the first batch of deliveries later this month.
“The Robocop can recognise people, register complaints, scan faces to match a suspect or missing-person description, for instance, even in public spaces like malls and offices,” says Psv . “It has 360-degree vision. It can chase a suspect, though laws don’t allow it to grab or arrest someone. It currently speaks and understands English, Hindi and Telugu.”
The Hyderabad police gave H-Bots feedback for the prototype. “Earlier, it was only supposed to be used to register complaints. They suggested we add a surveillance system and facial recognition. So we modified it accordingly. For test use in India, we asked them to restrict it to places like malls because of the rough terrain,” Psv says.
The company is also working on a humanoid for hospitals. “So instead of a nurse, a robot can record your body vitals like temperature and blood pressure, through a handshake, freeing up overworked nurses,” says Psv.
Domestically, Sheth of Ernst & Young says, there is potential for robotics deployment in hazardous jobs, such as in the Army.
The Defence Research and Development Organisation has already created a bot called Dakhs that can detect and destroy hazardous objects. Genrobotics, a Thiruvananthapuram-based company, has completed a test run of Bandicoot, a robot that cleans sewers. “It is in such areas that I see potential,” Sheth says.
So how many of the humanoids are likely to be deployed at home? Will we have Bandicoots on our streets and export batches of IRAs? Or will we too have entire hotels run by bots, as there already are in Japan? As a bot would tell you, in that stoic, flat-toned, mechanical voice —‘We do not know the answer to that question’.