Dozens of iconic sci-fi franchises, from Blade Runner to Terminator, center around the plot of realistic, humanoid robots so carefully built as to be practically indistinguishable from the real thing. Of course, we can’t build these machines yet — modern robots can do tricks and clean our floors, but they aren’t even a little bit human-like.
That said, we have begun to make progress and some of the early attempts at honest-to-god androids are pretty impressive… and just a little bit scary.
Below, we’ll count down some of the most impressive androids ever created. It’s nearly impossible to make an android demo that doesn’t look like the opening of a robot apocalypse movie, so buckle up: the future is coming, and it’s weirder than ever.
It would be impossible to make a list of impressive androids without mentioning ASIMO, the first android to capture peoples’ imaginations, and the first to introduce the idea of practical, humanoid robots to the popular consciousness. ASIMO, whose name stands for Advanced Step in Innovative MObility, is a three-foot-tall robot that resembles a cross between a space suit and an ipod.
On the hardware side, the platform features a number of sensors, an operating time of several hours on one charge, and a full 57 degrees of freedom (including thirteen in each hand). ASIMO can walk up and down stairs, pick up and manipulate objects, and was the first robot to demonstrate the ability to run (albeit at a sedate, waddling pace of 3.7 miles per hour).
On the software side, ASIMO comes with a variety of neat party tricks, including the ability to respond to voice commands, turn to look at noises, identify simple hand gestures, perform speech synthesis, and map its environment for navigation.
So why is ASIMO at the bottom of this list? Well, to put it succinctly, ASIMO is not a particularly robust robot. Watch this video:
The modern version of ASIMO has gone through many iterations, but is ultimately an incremental improvement on the original ASIMO — one of the first humanoid robots ever. As such, ASIMO is limited to the design paradigms available at the time, none of which can gracefully handle the unexpected (in this demo, the stairs were higher than ASIMO had been programmed to expect).
ASIMO fails when unexpected variables are introduced, which is why you tend to see the robot in very carefully structured public demonstrations. ASIMO uses a conservative style of walking, which minimizes the degree to which its center of gravity isn’t directly above its feet, an approach which led to ASIMO’s distinctive bent-knee duckstep. This, while simple to implement on a flat surface, fails on uneven terrain, and isn’t very fast, efficient, or natural-looking.
PETMAN (Protection Ensemble Test MANnequin), is the product of Boston Dynamics, a former DARPA contractor that was recently bought by Google, and who you may know from the famous BigDog video. PETMAN was designed to test hazmat suits for the military, making it probably one of the best examples of “feature creep” in human history. It takes a special kind of government contractor to go from testing hazmat suits to deciding to build a Terminator instead. In Boston Dynamics’ own words:
“PETMAN is an anthropomorphic robot designed for testing chemical protection clothing. Natural agile movement is essential for PETMAN to simulate how a soldier stresses protective clothing under realistic conditions. Unlike previous suit testers that had a limited repertoire of motion and had to be supported mechanically, PETMAN balances itself and moves freely; walking, bending and doing a variety of suit-stressing calisthenics during exposure to chemical warfare agents.”
PETMAN is impressive in many ways. While the machine lacks many of ASIMO’s features (like hands, vision, and skin) it does have impressive legs and extraordinary balance, allowing it to walk in a very human-like way across uneven terrain. The human stride is essentially a series of controlled falls, and PETMAN emulates it excellently.
As a consequence of this, PETMAN can walk and run quite a bit faster than ASIMO, and is capable of hiking over a variety of terrain. PETMAN also moves pneumatically, meaning that it can be powered by a gas engine, and can run autonomously for longer periods of time than conventional batteries would allow.
Before we move on, take a moment to watch that same video to the dulcet tones of the BeeGees. I’m not sure why that’s so funny, but it is.
NAO is a robot born out of the same philosophy as ASIMO: conservative walking, precise control, and many degrees of freedom. The difference is that NAO is much smaller, and designed to be an open, relatively inexpensive robotic platform. In other words, if you have $8,000 to blow on a hobby project, you can buy one right now.
NAO has a fairly broad suite of sensors, dexterous hands, and quite a bit of maneuverability. Plus, since anyone can write apps that run on NAO (which can be triggered by combinations of voice commands and visual data), the robot’s software continually gets smarter as time passes. NAO is the official platform of the robocup robot soccer competition. If you’re curious, robots are not particularly good at soccer.
The NAO bots have been used for a variety of tasks, including teaching social skills to autistic children, and providing an educational platform for robotics education. But, if you’re not excited by that, you can also make them dance to “Gangnam Style” and “Thriller“.
2. Martin Kelly’s Head
ASIMO, NAO, and PETMAN are all very impressive, but none of these robots look much like people. ASIMO and NAO are both the wrong size and proportion. PETMAN looks like a T-800 wearing track shoes. It has a warning light for a head, for goodness’ sakes! None of them have anything resembling a face. For good reason, too! A quick journey across Youtube will reveal the perils of trying to build plausible human faces – examples range from the merely dead-eyed and upsetting to genuine nightmare material.
These past failures make Martin Kelly’s example that much more impressive. The eyes seem alive, and the face moves naturally into a variety of convincing expressions. It almost seems like the robot knows something that we don’t. This is made all the more impressive by the relatively simple robotic rig that articulates the face. If you couldn’t see the table through the back of its open mouth, it would almost seem alive.
As it turns out, building and animating a plausible human face is more a question of artistic limitations than technological ones, and, as a result, turning artists with technical skill loose on the problem works much better than simply allowing robotics grad students to mess around with rubber and motors in a lab, which is presumably how the uncanny valley girl came into existence.
Following the success of PETMAN, Boston Dynamics continued to develop the core technology into a full-fledged robotics platform. The result is ATLAS, a sophisticated robot capable of moving through a variety of terrains and environments. ATLAS is designed for search and rescue, and has articulate hands, a full suite of sensors (including laser rangefinders and stereo cameras), and can balance very effectively.
ATLAS does, however, look quite a bit less human than PETMAN, due to the increased complexity of the hardware needed to operate. In fact, there’s something insect-like about its bulbous head and unblinking glass eyes. That said, ATLAS is incredibly impressive mechanically.
The premier of ATLAS caused one AI researcher, Gary Bradski, to comment:
“People love the wizards in Harry Potter or ‘Lord of the Rings,’ but this is real. […] A new species, Robo sapiens, are emerging,”
A DARPA official, Gill Pratt, was more pragmatic about the potential of the technology.
“Two weeks ago 19 brave firefighters lost their lives. […] A number of us who are in the robotics field see these events in the news, and the thing that touches us very deeply is a single kind of feeling which is, can’t we do better? All of this technology that we work on, can’t we apply that technology to do much better? I think the answer is yes.”
Following the Google buy-out, ATLAS belongs to Google now, which may eventually commit the technology to this kind of application. For now, though, you can still watch the robot’s training montage set to the Rocky theme.
The Robots of Tomorrow
Our ability to create robots and intelligent software is still very primitive, in many ways. Robotics is almost literally in its infancy: the hardware is just learning to walk properly and use fine motor control, while the software is finally starting to read at the level of an elementary-school student.
There’s a long way to go before machines can meet and exceed human capability. Still, progress is deceptively quick, and the odds are good that in your lifetime, you’ll continue being surprised by the leaps and bounds that machines make. The future of robotics is going to be interesting, and I, for one, am looking forward to it.
What is your favorite advanced humanoid robotics project? Do you think robots will start doing more complex human tasks in the near future? Share your thoughts in the comments section below!