The Future of Humanoid Autonomous Robots
Our Robotic Overlords’ may be a bit of an exaggeration when picturing the future of robotics, no doubt influenced by Hollywood. However, it is not an overstatement to imagine a near future where robots will be an essential part of our lives; living in our homes, working alongside factory workers and helping out in emergencies.
To achieve this ambitious future, researchers are focusing on a particular type of robot, referred to as humanoid.
What are Humanoid Robots?
A humanoid robot is distinctive in that it bears a shape resembling that of the human body. While some humanoid robots are admittedly more scary than human, the basic premise is that they have two legs, two arms, a head and a torso. Some robots, however, tend to take on a particular section of the human, such as the head, or from the waist up.
A popular term for humanoid robots, especially those designed to be aesthetically pleasing, is androids. This term is sometimes used to refer only to male-resembling robots while robots resembling human females are called Gynoids.
The most common reason for creating human robots in the human image is to ease their interaction with humans. Most of these humanoid robots being developed are designed to work in human environments such as homes and factories. A human-looking robot is certainly easier for people to accept than one that looks alien like.
Humanoid robots can also be developed for research purposes, to study various human aspects such as bipedal locomotion and the inner workings of the human body.
Impressive advances have already been made in this field. Organizations like Boston Dynamics, which was acquired by Google X for $500 million in 2013. Honda, which has already developed androids that can run, pick stuff up, open doors and even talk, and IBM’s think tank, Watson; capable of answering questions that are posed in natural language.
Even more interesting from what has already been achieved, is what lies ahead. How advanced can humanoid robots get? Can they, at some point, become perfect human mirrors in terms of appearance, function and even think? Will they one day become capable of emotion?
The best way to get answers to these questions is to look back at how far we have come, understand where we are today and predict what the future will look like.
How Far Have We Come?
The idea of creating automated machines, looking and functioning like humans or animals, goes as far back as the ancient cultures of China, Greece and Egypt.
One of the very first robots was created by a Greek inventor in-and-around 250 B.C. The apparatus was called Clepsydra, or a water clock.
Fast forward centuries later to one of the most brilliant intellects, the world has ever seen; Nikola Tesla. The man built the earliest robotic vehicles in the 1890s. Fast forward again, this time just a few years ahead to the 20th century where creations like Elsie the Tortoise, the Walking Truck by General Electric and the Unimates greatly helped advanced the field of robotics.
Humanoid robots have been an integral part of general robotic development. From early on, experts used the human body and all its intricacies as the template to create humanlike robots. One of the best-known early efforts to create a humanoid robot was by Leonardo Da Vinci. He created a robotic knight. Thought to have been created in 1495, the robotic knight could accomplish a number of lifelike actions, including standing, sitting and moving its arms.
The invention of modern humanoid robots kicked into high gear in the 1930s. In 1939, Westinghouse Electric Corporation showcased a robot that could smoke, move its head, wave its arms and blow balloons. In 1952, Ozamu Tezuka of Japan created the hugely popular Astroboy. In 1973, Waseda University of Tokyo created the first full scale humanoid robot that could talk, grip and transport objects.
Between 1996 and 2002, Honda introduced three highly advanced androids; the P2 in 1996, the P3 in 1997 and Asimo in 2002. Honda’s inventions, especially Asimo, represented a rapid advancement in robotics technology. Asimo, for example, was equipped with advanced artificial intelligence. It could walk, run, recognize faces and gestures, accept voice commands, follow faces during interaction and give a handshake.
Today, there are dozens of companies and organizations making their own creations. The result has been a proliferation of advanced humanoid robots with truly impressive features.
That being said, it would be a grave error not to emphasize the infancy of this field of robotics. Issues such as learning, locomotion and grip have yet to be perfected. We have just taken the first and second precarious steps; there are much more to go.
Humanoid Robots and Artificial Intelligence
The goal of humanoid robotics has always been to create a machine with the function and intelligence close to that of a human being. This means that the robot has to be able to function independently with little to no human intervention. In other words, the robot should be programmable.
Early humanoid robots were non-programmable. They were operated through an electronic circuitry and some earlier ones even used air pressure for manipulation.
The first programmable robot was created in 1954 by George Devol. He named it the Unimate. It was used by General Motors in their manufacturing process.
Over the years, humanoid robots have increasingly possessed sophisticated programming, allowing them to perform a wide range of tasks autonomously. This programming, or artificial intelligence, has been getting more advanced with time.
The earliest form of artificial intelligence was modeled after insects. It is referred to as Nouvelle artificial intelligence. Pioneered by Rodney Brooks in the 1980s, this basic form of AI was limited to the intelligence level of insects. Though quite intelligent, insect-level AI proved severely limiting. Insectoid robots based on Nouvelle’s AI had limited features, such as collision avoidance and basic locomotion.
Rodney Brooks, the same man who pioneered Nouvelle’s AI, made a sharp pivot in the 1990s. He decided to pursue human-level AI, creating a humanoid robot called Cog. This was the beginning of the use of advanced artificial intelligence in humanoid robots.
Today, androids are still a long way from human-level intelligence, but they have made huge leaps. Humanoid robots can now take sensory information from their surroundings, process it quickly, make a wide array of autonomous decisions based on it, and most importantly, learn from it.
The future of AI, as we are going to see shortly, is perhaps the most exciting part about humanoid robotics. The all-encompassing question is; how intelligent can robots get?
Continually pushing the intelligence limit is ultimately the future for humanoid robotics.
The Future of Humanoid Robotics
We have barely scratched the surface of humanoid robotics. Sure, we have seen some marvelous creations and credit is due to the minds behind these great inventions. But in terms of potential, there is much yet to be achieved. This is not meant as a discouragement but rather a promise that the best is yet to come.
Humanoid robotics is gathering pace, with plenty of money and skill being thrown into the field, and you can expect research and development to pick up the pace. It would not be overly optimistic to expect a fully functional android, with a high level of intelligence, in two decades or less.
In looking forward to the future of humanoid robots, there are three main areas worth discussing. They include:
• The challenges blocking the path to that future
• The possibilities the future holds
• The risks that come with that future
1. The Challenges
i. Bipedal Locomotion
As we noted in the beginning, having two legs is an unspoken standard for designing humanoid robots. By extension therefore, these robots must portray bipedal locomotion just like humans. While walking may come easy for you and me, the mechanics of walking is quite complicated. Replicating these mechanics in robots has proven to be a major challenge. Although, small humanoid robots have seen some significant success in this area, thanks to their large feet and low center of gravity. Larger robots such as Atlas by Boston Dynamics end up looking awkwardly insect-like when their center of gravity is lowered. Their bent knees and lumbering gait do not make for a pleasant sight. As shown in this video of a robot competition hosted by DARPA, most large humanoid robots cannot even stay up for long.
The challenge is to create a humanoid robot that does not fall over, has perfect balance even on rough terrain and can achieve significant walking and running speeds.
ii. Human-level artificial intelligence
In 2011, an intelligent computing system developed by IBM won against two human contestants in the game show, Jeopardy!’. It won the first-place grand prize of one million dollars. The system was named, Watson, after IBM’s first CEO, and is the epitome of the rapid advancement of artificial intelligence.
IBM Watson relies on natural language processing to process questions posed to it and responds in natural language. Through machine learning, the system quickly learns about a certain subject from materials such as PDFs, Ebooks and web pages and gets better with time; as its users keep asking questions.
To prove its abilities, IBM Watson has posed as a teacher’s assistant for an online course. By the end of the semester, students were shocked to discover that the Jill Watson they were emailing and asking queries online was actually an AI. Even more impressive, IBM has deployed its system to work as a lawyer in a top law firm in New York. It will be handling bankruptcy issues; answering questions by human lawyers and keeping them updated on the latest legal issues and ongoing cases affecting their casework.
The development of artificial intelligence is closely tied to the success of the next age of humanoid robots. IBM’s Watson give us hope that a fantasy future is closer to reality than we think. But there is much to be done. Watson has to be trained extensively in a specific field before it can be launched to work within it. What we need is AI, smart enough to learn anything and everything on its own and use that knowledge to interact successfully with humans.
Like bipedal locomotion, this is another area that requires a lot of improvement if humanoid robots are to be successfully integrated into human environments. How well a robot can manipulate its surroundings, particularly through arms and hands, is an important measure of its overall success.
We have made some significant gains in this area. Through the use of fine motor control, cameras and tactile sensors, humanoid robots can pick up objects, carry them for a distance and place them back down. Some robots can even open doors, wield tools such as a hammer and offer up a handshake when it meets a person.
But when it comes to more intricate actions such as holding delicate objects and using tools such as knives, robots fails. Again, mechanics is the central issue here. The human hand is very complex with over 20 degrees of freedom. To avoid mechanical complexity, most robots have between 5 and 7 degrees of freedom in their arms, thus sacrificing a fine mastery and efficiency in arm movement.
Other than the hands and arms, proper manipulation also encompasses whole-body movement. For instance, a robot bending to pick up a package will have to engage their entire body while maintaining balance.
Today’s humanoid robots exhibit a high degree of manipulation only when in tightly controlled environments. Tomorrow’s robots will need the ability to adapt quickly to deal with ever-changing surroundings.
When we are finally able to achieve perfect or at least drastically improved manipulation in humanoid robots, we can deploy them to a wide range of applications. They can work efficiently in the kitchen, assist surgeons, rescue people quickly in catastrophes and perhaps even join the army; though with ethical implications.
One defining characteristic of humans is their ability to learn throughout their lives. Experts have been trying to emulate this ability in humanoid robots by using advanced AI. Like so many other aspects of humanoid robotics, we are still at the elementary level.
Currently, there are a number of ways that robots are learning. One of them is through imitation. By observing human actions, robots could replicate their motions and even perfect them. Another technique is experiential learning. Here, humanoid robots become better the longer they are immersed in a specific environment. Their experiences become ingrained lessons; a good example of this is IBM Watson.
A third method is trial and error learning. This is where a robot is directed to repeatedly handle a task until it succeeds. A great example of this is a study at Carnegie Mellon University where a robot was left alone in a room with various objects. Through repeated tries, and plenty of errors, the robot learnt to grasp properly in 10 days.
It is obvious that we already have a ‘grasp’ on robot learning. All that is left now is to fine tune these techniques to allow for faster, better and more varied learning. It is especially important for robots to be able to learn quickly within a varying environment, an ability referred to as adaptive control.
2. The Possibilities
i. The near-perfect humanoid robot
Can you imagine a world where all the above challenges have been solved? It will be a world where we finally get a near-perfect humanoid robot. This will be a robot that can read human emotions and respond accordingly, have almost human-level intelligence, almost indistinguishable from another human during a conversation, and will achieve perfect human-like motion and manipulation.
Advances in robotics are something of a snowball effect; it started out slow and precarious but has gained pace in recent years. So expect more developments in the next few years than there have been in the last century.
In the next two or three decades, humanoid robotics will be leaps ahead of what is currently possible.
ii. Numerous applications
As humanoid robots become better in various aspects — dexterity, motion, audio and visual perception, human interaction and so on — we will find uses for them in a wide variety of fields. Case in point:
Stanford has an underwater humanoid robot diver, OceanOne, used to study coral reefs deep in the sea.
NASA has a humanoid robot, Valkyrie, currently receiving training in preparation for sending it to mars.
Honda is preparing a new version of ASIMO to deal with disaster situations.
In Brooklyn NAO robots entertained the crowd with a ballet number.
Tiny humanoid robots are currently flying planes in simulators in preparation of flying real ones.
DARPA has long been involved in efforts to create a humanoid robot that can be used for military purposes. There are so many more applications that will benefit from the development of humanoid robotics. In the next few decades, you could be interacting with a robot surgeon, housekeeper or just even for companionship.
iii. Mass-market production
There are companies already producing consumer humanoid robots, though most of them feature only basic capabilities. But even these will cost you more than $700, with most costing over $1,000. Some high-end robots can cost you as much as $10,000.
Due to cost prohibitions, companies like Boston Dynamics and Honda cannot mass produce their advanced robots. This leaves us with only the basic options available from a number of sellers. But in two decades or so, this will change. As better humanoid robots are created, they will be made available to the public, first at high prices but gradually at lower price tags.
Within a few decades, humanoid robots will become some of the most popular consumer products, with numerous options available at different price ranges.
3. The Risks
It would be unfair to talk about the future of humanoid robots without considering whether there are any downsides to such a future. While a robot revolt might be a distant possibility, various other risks need to be taken into consideration, even as we chart new territories in this area.
One obvious risk that has ignited heated debate in many circles is that of the potential economic impact. It is highly likely that humanoid robots will take over many jobs currently being done by humans. We already have robots working in factories. But the biggest impact will likely be felt in the service industry. Robots will replace hotel workers, waiters, store attendants and customer service agents.
With all these jobs disappearing, there is serious need to discuss measures to mitigate any potential economic crisis. One idea that has been floated around is having a basic guaranteed wage for every United States citizen.
For humanoid robots used in areas such as the military and medicine, safety is a major issue. Even the best of machines are not a hundred percent foolproof.
For such a novel concept that has just started taking root, it is really hard to predict all the problems that could arise. The only option that remains is to tackle challenges as we meet them.
Major Industry Players
We have exhaustively dissected the future of humanoid robotics, detailing what challenges, possibilities and risks lay ahead. To wrap it up, it is important to know the people and organizations driving the new age.
• Boston Dynamics — this is one of the leading makers of advanced robots, having once been a part of MIT. It was acquired by Google X in 2013 though it is already being put up for sale. Boston Dynamics created popular inventions such as Atlas, Big Dog and Petman.
• Google — with the potential parting of Google and Boston Dynamics, Google will focus more on the various other ventures it runs. One of them involves artificial intelligence, similar to IBM Watson. In 2016, AlphaGo (a Google AI) beat a professional Go Player 4–1 in a 5-match competition.
• Honda Robotics — Honda is popular for its ASIMO lineup. ASIMO is one of the most advanced humanoid robots currently available.
• IBM — IBM Watson has already shown what it can do. It has gone from mere experiments into actually helping businesses and institutions in their operations.
Other smaller companies involved in humanoid robotics include Kawada Robotics, Meka Robotics, and AvatarMind.
If you look at all the technological developments being made today — artificial intelligence, driverless cars, internet of things — they all point towards a smarter future. Whether we are ready or not, humanoid robots will be an unavoidable part of that future. They will change how we work, play and interact. The world is hurtling toward a new age, one that will be better for businesses, organizations, communities, families, and individuals. Get ready.
Originally published at SiNGLE GADGET.