Tag Archives: jump

#435757 Robotic Animal Agility

Posted on September 19, 2019 by Android

An off-shore wind power platform, somewhere in the North Sea, on a freezing cold night, with howling winds and waves crashing against the impressive structure. An imperturbable ANYmal is quietly conducting its inspection.

ANYmal, a medium sized dog-like quadruped robot, walks down the stairs, lifts a “paw” to open doors or to call the elevator and trots along corridors. Darkness is no problem: it knows the place perfectly, having 3D-mapped it. Its laser sensors keep it informed about its precise path, location and potential obstacles. It conducts its inspection across several rooms. Its cameras zoom in on counters, recording the measurements displayed. Its thermal sensors record the temperature of machines and equipment and its ultrasound microphone checks for potential gas leaks. The robot also inspects lever positions as well as the correct positioning of regulatory fire extinguishers. As the electronic buzz of its engines resumes, it carries on working tirelessly.

After a little over two hours of inspection, the robot returns to its docking station for recharging. It will soon head back out to conduct its next solitary patrol. ANYmal played alongside Mulder and Scully in the “X-Files” TV series*, but it is in no way a Hollywood robot. It genuinely exists and surveillance missions are part of its very near future.

Off-shore oil platforms, the first test fields and probably the first actual application of ANYmal. ©ANYbotics

This quadruped robot was designed by ANYbotics, a spinoff of the Swiss Federal Institute of Technology in Zurich (ETH Zurich). Made of carbon fibre and aluminium, it weighs about thirty kilos. It is fully ruggedised, water- and dust-proof (IP-67). A kevlar belly protects its main body, carrying its powerful brain, batteries, network device, power management system and navigational systems.

ANYmal was designed for all types of terrain, including rubble, sand or snow. It has been field tested on industrial sites and is at ease with new obstacles to overcome (and it can even get up after a fall). Depending on its mission, its batteries last 2 to 4 hours.

On its jointed legs, protected by rubber pads, it can walk (at the speed of human steps), trot, climb, curl upon itself to crawl, carry a load or even jump and dance. It is the need to move on all surfaces that has driven its designers to choose a quadruped. “Biped robots are not easy to stabilise, especially on irregular terrain” explains Dr Péter Fankhauser, co-founder and chief business development officer of ANYbotics. “Wheeled or tracked robots can carry heavy loads, but they are bulky and less agile. Flying drones are highly mobile, but cannot carry load, handle objects or operate in bad weather conditions. We believe that quadrupeds combine the optimal characteristics, both in terms of mobility and versatility.”

What served as a source of inspiration for the team behind the project, the Robotic Systems Lab of the ETH Zurich, is a champion of agility on rugged terrain: the mountain goat. “We are of course still a long way” says Fankhauser. “However, it remains our objective on the longer term.

The first prototype, ALoF, was designed already back in 2009. It was still rather slow, very rigid and clumsy – more of a proof of concept than a robot ready for application. In 2012, StarlETH, fitted with spring joints, could hop, jump and climb. It was with this robot that the team started participating in 2014 in ARGOS, a full-scale challenge, launched by the Total oil group. The idea was to present a robot capable of inspecting an off-shore drilling station autonomously.

Up against dozens of competitors, the ETH Zurich team was the only team to enter the competition with such a quadrupedal robot. They didn’t win, but the multiple field tests were growing evermore convincing. Especially because, during the challenge, the team designed new joints with elastic actuators made in-house. These joints, inspired by tendons and muscles, are compact, sealed and include their own custom control electronics. They can regulate joint torque, position and impedance directly. Thanks to this innovation, the team could enter the same competition with a new version of its robot, ANYmal, fitted with three joints on each leg.

The ARGOS experience confirms the relevance of the selected means of locomotion. “Our robot is lighter, takes up less space on site and it is less noisy” says Fankhauser. “It also overcomes bigger obstacles than larger wheeled or tracked robots!” As ANYmal generated public interest and its transformation into a genuine product seemed more than possible, the startup ANYbotics was launched in 2016. It sold not only its robot, but also its revolutionary joints, called ANYdrive.

Today, ANYmal is not yet ready for sale to companies. However, ANYbotics has a growing number of partnerships with several industries, testing the robot for a few days or several weeks, for all types of tasks. Last October, for example, ANYmal navigated its way through the dark sewage system of the city of Zurich in order to test its capacity to help workers in similar difficult, repetitive and even dangerous tasks.

Why such an early interest among companies? “Because many companies want to integrate robots into their maintenance tasks” answers Fankhauser. “With ANYmal, they can actually evaluate its feasibility and plan their strategy. Eventually, both the architecture and the equipment of buildings could be rethought to be adapted to these maintenance robots”.

ANYmal requires ruggedised, sealed and extremely reliable interconnection solutions, such as LEMO. ©ANYbotics

Through field demonstrations and testing, ANYbotics can gather masses of information (up to 50,000 measurements are recorded every second during each test!) “It helps us to shape the product.” In due time, the startup will be ready to deliver a commercial product which really caters for companies’ needs.

Inspection and surveillance tasks on industrial sites are not the only applications considered. The startup is also thinking of agricultural inspections – with its onboard sensors, ANYmal is capable of mapping its environment, measuring bio mass and even taking soil samples. In the longer term, it could also be used for search and rescue operations. By the way, the robot can already be switched to “remote control” mode at any time and can be easily tele-operated. It is also capable of live audio and video transmission.

The transition from the prototype to the marketed product stage will involve a number of further developments. These include increasing ANYmal’s agility and speed, extending its capacity to map large-scale environments, improving safety, security, user handling and integrating the system with the customer’s data management software. It will also be necessary to enhance the robot’s reliability “so that it can work for days, weeks, or even months without human supervision.” All required certifications will have to be obtained. The locomotion system, which had triggered the whole business, is only one of a number of considerations of ANYbotics.

Designed for extreme environments, for ANYmal smoke is not a problem and it can walk in the snow, through rubble or in water. ©ANYbotics

The startup is not all alone. In fact, it has sold ANYmal robots to a dozen major universities who use them to develop their know-how in robotics. The startup has also founded ANYmal Research, a community including members such as Toyota Research Institute, the German Aerospace Center and the computer company Nvidia. Members have full access to ANYmal’s control software, simulations and documentation. Sharing has boosted both software and hardware ideas and developments (built on ROS, the open-source Robot Operating System). In particular, payload variations, providing for expandability and scalability. For instance, one of the universities uses a robotic arm which enables ANYmal to grasp or handle objects and open doors.

Among possible applications, ANYbotics mentions entertainment. It is not only about playing in more films or TV series, but rather about participating in various attractions (trade shows, museums, etc.). “ANYmal is so novel that it attracts a great amount of interest” confirms Fankhauser with a smile. “Whenever we present it somewhere, people gather around.”

Videos of these events show a fascinated and sometimes slightly fearful audience, when ANYmal gets too close to them. Is it fear of the “bad robot”? “This fear exists indeed and we are happy to be able to use ANYmal also to promote public awareness towards robotics and robots.” Reminiscent of a young dog, ANYmal is truly adapted for the purpose.

However, Péter Fankhauser softens the image of humans and sophisticated robots living together. “These coming years, robots will continue to work in the background, like they have for a long time in factories. Then, they will be used in public places in a selective and targeted way, for instance for dangerous missions. We will need to wait another ten years before animal-like robots, such as ANYmal will share our everyday lives!”

At the Consumer Electronics Show (CES) in Las Vegas in January, Continental, the German automotive manufacturing company, used robots to demonstrate a last-mile delivery. It showed ANYmal getting out of an autonomous vehicle with a parcel, climbing onto the front porch, lifting a paw to ring the doorbell, depositing the parcel before getting back into the vehicle. This futuristic image seems very close indeed.

*X-Files, season 11, episode 7, aired in February 2018 Continue reading →

Posted in Human Robots

#435752 T-RHex Is a Hexapod Robot With ...

Posted on September 19, 2019 by Android

In Aaron Johnson’s “Robot Design & Experimentation” class at CMU, teams of students have a semester to design and build an experimental robotic system based on a theme. For spring 2019, that theme was “Bioinspired Robotics,” which is definitely one of our favorite kinds of robotics—animals can do all kinds of crazy things, and it’s always a lot of fun watching robots try to match them. They almost never succeed, of course, but even basic imitation can lead to robots with some unique capabilities.

One of the projects from this year’s course, from Team ScienceParrot, is a new version of RHex called T-RHex (pronounced T-Rex, like the dinosaur). T-RHex comes with a tail, but more importantly, it has tiny tapered toes, which help it grip onto rough surfaces like bricks, wood, and concrete. It’s able to climb its way up very steep slopes, and hang from them, relying on its toes to keep itself from falling off.

T-RHex’s toes are called microspines, and we’ve seen them in all kinds of robots. The most famous of these is probably JPL’s LEMUR IIB (which wins on sheer microspine volume), although the concept goes back at least 15 years to Stanford’s SpinyBot. Robots that use microspines to climb tend to be fairly methodical at it, since the microspines have to be engaged and disengaged with care, limiting their non-climbing mobility.

T-RHex manages to perform many of the same sorts of climbing and hanging maneuvers without losing RHex’s ability for quick, efficient wheel-leg (wheg) locomotion.

If you look closely at T-RHex walking in the video, you’ll notice that in its normal forward gait, it’s sort of walking on its ankles, rather than its toes. This means that the microspines aren’t engaged most of the time, so that the robot can use its regular wheg motion to get around. To engage the microspines, the robot moves its whegs backwards, meaning that its tail is arguably coming out of its head. But since all of T-RHex’s capability is mechanical in nature and it has no active sensors, it doesn’t really need a head, so that’s fine.

The highest climbable slope that T-RHex could manage was 55 degrees, meaning that it can’t, yet, conquer vertical walls. The researchers were most surprised by the robot’s ability to cling to surfaces, where it was perfectly happy to hang out on a slope of 135 degrees, which is a 45 degree overhang (!). I have no idea how it would ever reach that kind of position on its own, but it’s nice to know that if it ever does, its spines will keep doing their job.

Photo: CMU

T-RHex uses laser-cut acrylic legs, with the microspines embedded into 3D-printed toes. The tail is needed to prevent the robot from tipping backward.

For more details about the project, we spoke with Team ScienceParrot member (and CMU PhD student) Catherine Pavlov via email.

IEEE Spectrum: We’re used to seeing RHex with compliant, springy legs—how do the new legs affect T-RHex’s mobility?

Catherine Pavlov: There’s some compliance in the legs, though not as much as RHex—this is driven by the use of acrylic, which was chosen for budget/manufacturing reasons. Matching the compliance of RHex with acrylic would have made the tines too weak (since often only a few hold the load of the robot during climbing). It definitely means you can’t use energy storage in the legs the way RHex does, for example when pronking. T-RHex is probably more limited by motor speed in terms of mobility though. We were using some borrowed Dynamixels that didn’t allow for good positioning at high speeds.

How did you design the climbing gait? Why not use the middle legs, and why is the tail necessary?

The gait was a lot of hand-tuning and trial-and-error. We wanted a left/right symmetric gait to enable load sharing among more spines and prevent out-of-plane twisting of the legs. When using all three pairs, you have to have very accurate angular positioning or one leg pair gets pushed off the wall. Since two legs should be able to hold the full robot gait, using the middle legs was hurting more than it was helping, with the middle legs sometimes pushing the rear ones off of the wall.

The tail is needed to prevent the robot from tipping backward and “sitting” on the wall. During static testing we saw the robot tip backward, disengaging the front legs, at around 35 degrees incline. The tail allows us to load the front legs, even when they’re at a shallow angle to the surface. The climbing gait we designed uses the tail to allow the rear legs to fully recirculate without the robot tipping backward.

Photo: CMU

Team ScienceParrot with T-RHex.

What prevents T-RHex from climbing even steeper surfaces?

There are a few limiting factors. One is that the tines of the legs break pretty easily. I think we also need a lighter platform to get fully vertical—we’re going to look at MiniRHex for future work. We’re also not convinced our gait is the best it can be, we can probably get marginal improvements with more tuning, which might be enough.

Can the microspines assist with more dynamic maneuvers?

Dynamic climbing maneuvers? I think that would only be possible on surfaces with very good surface adhesion and very good surface strength, but it’s certainly theoretically possible. The current instance of T-RHex would definitely break if you tried to wall jump though.

What are you working on next?

Our main target is exploring the space of materials for leg fabrication, such as fiberglass, PLA, urethanes, and maybe metallic glass. We think there’s a lot of room for improvement in the leg material and geometry. We’d also like to see MiniRHex equipped with microspines, which will require legs about half the scale of what we built for T-RHex. Longer-term improvements would be the addition of sensors e.g. for wall detection, and a reliable floor-to-wall transition and dynamic gait transitions.

[ T-RHex ] Continue reading →

Posted in Human Robots

#435738 Boing Goes the Trampoline Robot

Posted on September 18, 2019 by Android

There are a handful of quadrupedal robots out there that are highly dynamic, with the ability to run and jump, but those robots tend to be rather expensive and complicated, requiring powerful actuators and legs with elasticity. Boxing Wang, a Ph.D. student in the College of Control Science and Engineering at Zhejiang University in China, contacted us to share a project he’s been working to investigate quadruped jumping with simple, affordable hardware.

“The motivation for this project is quite simple,” Boxing says. “I wanted to study quadrupedal jumping control, but I didn’t have custom-made powerful actuators, and I didn’t want to have to design elastic legs. So I decided to use a trampoline to make a normal servo-driven quadruped robot to jump.”

Boxing and his colleagues had wanted to study quadrupedal running and jumping, so they built this robot with the most powerful servos they had access to: Kondo KRS6003RHV actuators, which have a maximum torque of 6 Nm. After some simple testing, it became clear that the servos were simply not fast or powerful enough to get the robot to jump, and that an elastic element was necessary to store energy to help the robot get off the ground.

“Normally, people would choose elastic legs,” says Boxing. “But nobody in my lab knew for sure how to design them. If we tried making elastic legs and we failed to make the robot jump, we couldn’t be sure whether the problem was the legs or the control algorithms. For hardware, we decided that it’s better to start with something reliable, something that definitely won’t be the source of the problem.”

As it turns out, all you need is a trampoline, an inertial measurement unit (IMU), and little tactile switches on the end of each foot to detect touch-down and lift-off events, and you can do some useful jumping research without a jumping robot. And the trampoline has other benefits as well—because it’s stiffer at the edges than at the center, for example, the robot will tend to center itself on the trampoline, and you get some warning before things go wrong.

“I can’t say that it’s a breakthrough to make a quadruped robot jump on a trampoline,” Boxing tells us. “But I believe this is useful for prototype testing, especially for people who are interested in quadrupedal jumping control but without a suitable robot at hand.”

To learn more about the project, we emailed him some additional questions.

IEEE Spectrum: Where did this idea come from?

Boxing Wang: The idea of the trampoline came while we were drinking milk tea. I don’t know why it came up, maybe someone saw a trampoline in a gym recently. And I don’t remember who proposed it exactly. It was just like someone said it unintentionally. But I realized that a trampoline would be a perfect choice. It’s reliable, easy to buy, and should have a similar dynamic model with the one of jumping with springy legs (we have briefly analyzed this in a paper). So I decided to try the trampoline.

How much do you think you can learn using a quadruped on a trampoline, instead of using a jumping quadruped?

Generally speaking, no contact surfaces are strictly rigid. They all have elasticity. So there are no essential differences between jumping on a trampoline and jumping on a rigid surface. However, using a quadruped on a trampoline can give you more information on how to make use of elasticity to make jumping easier and more efficient. You can use quadruped robots with springy legs to address the same problem, but that usually requires much more time on hardware design.

We prefer to treat the trampoline experiment as a kind of early test for further real jumping quadruped design. Unless you’re interested in designing an acrobatic robot on a trampoline, a real jumping quadruped is probably a more useful application, and that is our ultimate goal. The point of the trampoline tests is to develop the control algorithms first, and to examine the stability of the general hardware structure. Due to the similarity between jumping on a trampoline with rigid legs and jumping on hard surfaces with springy legs, the control algorithms you develop could be transferred to hard-surface jumping robots.

“Unless you’re interested in designing an acrobatic robot on a trampoline, a real jumping quadruped is probably a more useful application, and that is our ultimate goal. The point of the trampoline tests is to develop the control algorithms first, and to examine the stability of the general hardware structure”

Do you think that this idea can be beneficial for other kinds of robotics research?

Yes. For jumping quadrupeds with springy legs, the control algorithms could be first designed through trampoline tests using simple rigid legs. And the hardware design for elastic legs could be accelerated with the help of the control algorithms you design. In addition, we believe our work could be a good example of using a position-control robot to realize dynamic motions such as jumping, or even running.

Unlike other dynamic robots, every active joint in our robot is controlled through commercial position-control servos and not custom torque control motors. Most people don’t think that a position-control robot could perform highly dynamic motions such as jumping, because position-control motors usually mean high a gear ratio and slow response. However, our work indicates that, with the help of elasticity, stable jumping could be realized through position-control servos. So for those who already have a position-control robot at hand, they could explore the potential of their robot through trampoline tests.

Why is teaching a robot to jump important?

There are many scenarios where a jumping robot is needed. For example, a real jumping quadruped could be used to design a running quadruped. Both experience moments when all four legs are in the air, and it is easier to start from jumping and then move to running. Specifically, hopping or pronking can easily transform to bounding if the pitch angle is not strictly controlled. A bounding quadruped is similar to a running rabbit, so for now it can already be called a running quadruped.

To the best of our knowledge, a practical use of jumping quadrupeds could be planet exploration, just like what SpaceBok was designed for. In a low-gravity environment, jumping is more efficient than walking, and it’s easier to jump over obstacles. But if I had a jumping quadruped on Earth, I would teach it to catch a ball that I throw at it by jumping. It would be fantastic!

That would be fantastic.

Since the whole point of the trampoline was to get jumping software up and running with a minimum of hardware, the next step is to add some springy legs to the robot so that the control system the researchers developed can be tested on hard surfaces. They have a journal paper currently under revision, and Boxing Wang is joined as first author by his adviser Chunlin Zhou, undergrads Ziheng Duan and Qichao Zhu, and researchers Jun Wu and Rong Xiong. Continue reading →

Posted in Human Robots

#435733 Robot Squid and Robot Scallop Showcase ...

Posted on September 18, 2019 by Android

Most underwater robots use one of two ways of getting around. Way one is with propellers, and way two is with fins. But animals have shown us that there are many more kinds of underwater locomotion, potentially offering unique benefits to robots. We’ll take a look at two papers from ICRA this year that showed bioinspired underwater robots moving in creative new ways: A jet-powered squid robot that can leap out of the water, plus a robotic scallop that moves just like the real thing.

Image: Beihang University

Prototype of the squid robot in (a) open and (b) folded states. The soft fins and arms are controlled by pneumatic actuators.

This “squid-like aquatic-aerial vehicle” from Beihang University in China is modeled after flying squids. Real squids, in addition to being tasty, propel themselves using water jets, and these jets are powerful enough that some squids can not only jump out of the water, but actually achieve controlled flight for a brief period by continuing to jet while in the air. The flight phase is extended through the use of fins as arms and wings to generate a little bit of lift. Real squids use this multimodal propulsion to escape predators, and it’s also much faster—a squid can double its normal swimming speed while in the air, flying at up to 50 body lengths per second.

The squid robot is powered primarily by compressed air, which it stores in a cylinder in its nose (do squids have noses?). The fins and arms are controlled by pneumatic actuators. When the robot wants to move through the water, it opens a value to release a modest amount of compressed air; releasing the air all at once generates enough thrust to fire the robot squid completely out of the water.

The jumping that you see at the end of the video is preliminary work; we’re told that the robot squid can travel between 10 and 20 meters by jumping, whereas using its jet underwater will take it just 10 meters. At the moment, the squid can only fire its jet once, but the researchers plan to replace the compressed air with something a bit denser, like liquid CO2, which will allow for extended operation and multiple jumps. There’s also plenty of work to do with using the fins for dynamic control, which the researchers say will “reveal the superiority of the natural flying squid movement.”

“Design and Experiments of a Squid-like Aquatic-aerial Vehicle With Soft Morphing Fins and Arms,” by Taogang Hou, Xingbang Yang, Haohong Su, Buhui Jiang, Lingkun Chen, Tianmiao Wang, and Jianhong Liang from Beihang University in China, was presented at ICRA 2019 in Montreal.

Image: EPFL

The EPFL researchers studied the morphology and function of a real scallop (a) to design their robot scallop (b), which consists of two shells connected at a hinge and enclosed by a flexible elastic membrane. The robot and animal both swim by rapidly, cyclicly opening and closing their shells to generate water jets for propulsion. When the robot shells open, water is drawn into the body through rear openings near the hinge. When the shells close rapidly, the water is forced out, propelling the robot forward (c).

RoboScallop, a “bivalve inspired swimming robot,” comes from EPFL’s Reconfigurable Robotics Laboratory, headed by Jamie Paik. Real scallops, in addition to being tasty, propel themselves by opening and closing their shells to generate jets of water out of their backsides. By repetitively opening their shells slowly and then closing quickly, scallops can generate forward thrust in a way that’s completely internal to their bodies. Relative to things like fins or spinning propellers, a scallop is simple and robust, especially as you scale down or start looking at large swarms of robots. The EPFL researchers describe their robotic scallop as representing “a unique combination of robust to hazards or sustained use, safe in delicate environments, and simple by design.”

And here’s how the real thing looks:

As you can see from the video, RoboScallop is safe to handle even while it’s operating, although a gentle nibbling is possible if you get too handsy with it. Since the robot sucks water in and then jets it out immediately, the design is resistant to fouling, which can be a significant problem in marine environments. The RoboScallop prototype weighs 65 grams, and tops out at a brisk 16 centimeters per second, while clapping (that’s the actual technical) at just over 2.5 Hz. While RoboScallop doesn’t yet steer, real scallops can change direction by jetting out more water on one side than the other, and RoboScallop should be able to do this as well. The researchers also suggest that RoboScallop itself could even double as a gripper, which as far as I know, is not something that real scallops can do.

“RoboScallop: A Bivalve-Inspired Swimming Robot,” by Matthew A. Robertson, Filip Efremov, and Jamie Paik, was presented at ICRA 2019 in Montreal. Continue reading →

Posted in Human Robots

#435707 AI Agents Startle Researchers With ...

Posted on September 17, 2019 by Android

After 25 million games, the AI agents playing hide-and-seek with each other had mastered four basic game strategies. The researchers expected that part.

After a total of 380 million games, the AI players developed strategies that the researchers didn’t know were possible in the game environment—which the researchers had themselves created. That was the part that surprised the team at OpenAI, a research company based in San Francisco.

The AI players learned everything via a machine learning technique known as reinforcement learning. In this learning method, AI agents start out by taking random actions. Sometimes those random actions produce desired results, which earn them rewards. Via trial-and-error on a massive scale, they can learn sophisticated strategies.

In the context of games, this process can be abetted by having the AI play against another version of itself, ensuring that the opponents will be evenly matched. It also locks the AI into a process of one-upmanship, where any new strategy that emerges forces the opponent to search for a countermeasure. Over time, this “self-play” amounted to what the researchers call an “auto-curriculum.”

According to OpenAI researcher Igor Mordatch, this experiment shows that self-play “is enough for the agents to learn surprising behaviors on their own—it’s like children playing with each other.”

Reinforcement is a hot field of AI research right now. OpenAI’s researchers used the technique when they trained a team of bots to play the video game Dota 2, which squashed a world-champion human team last April. The Alphabet subsidiary DeepMind has used it to triumph in the ancient board game Go and the video game StarCraft.

Aniruddha Kembhavi, a researcher at the Allen Institute for Artificial Intelligence (AI2) in Seattle, says games such as hide-and-seek offer a good way for AI agents to learn “foundational skills.” He worked on a team that taught their AllenAI to play Pictionary with humans, viewing the gameplay as a way for the AI to work on common sense reasoning and communication. “We are, however, quite far away from being able to translate these preliminary findings in highly simplified environments into the real world,” says Kembhavi.

Illustration: OpenAI

AI agents construct a fort during a hide-and-seek game developed by OpenAI.

In OpenAI’s game of hide-and-seek, both the hiders and the seekers received a reward only if they won the game, leaving the AI players to develop their own strategies. Within a simple 3D environment containing walls, blocks, and ramps, the players first learned to run around and chase each other (strategy 1). The hiders next learned to move the blocks around to build forts (2), and then the seekers learned to move the ramps (3), enabling them to jump inside the forts. Then the hiders learned to move all the ramps into their forts before the seekers could use them (4).

The two strategies that surprised the researchers came next. First the seekers learned that they could jump onto a box and “surf” it over to a fort (5), allowing them to jump in—a maneuver that the researchers hadn’t realized was physically possible in the game environment. So as a final countermeasure, the hiders learned to lock all the boxes into place (6) so they weren’t available for use as surfboards.

Illustration: OpenAI

An AI agent uses a nearby box to surf its way into a competitor’s fort.

In this circumstance, having AI agents behave in an unexpected way wasn’t a problem: They found different paths to their rewards, but didn’t cause any trouble. However, you can imagine situations in which the outcome would be rather serious. Robots acting in the real world could do real damage. And then there’s Nick Bostrom’s famous example of a paper clip factory run by an AI, whose goal is to make as many paper clips as possible. As Bostrom told IEEE Spectrum back in 2014, the AI might realize that “human bodies consist of atoms, and those atoms could be used to make some very nice paper clips.”

Bowen Baker, another member of the OpenAI research team, notes that it’s hard to predict all the ways an AI agent will act inside an environment—even a simple one. “Building these environments is hard,” he says. “The agents will come up with these unexpected behaviors, which will be a safety problem down the road when you put them in more complex environments.”

AI researcher Katja Hofmann at Microsoft Research Cambridge, in England, has seen a lot of gameplay by AI agents: She started a competition that uses Minecraft as the playing field. She says the emergent behavior seen in this game, and in prior experiments by other researchers, shows that games can be a useful for studies of safe and responsible AI.

“I find demonstrations like this, in games and game-like settings, a great way to explore the capabilities and limitations of existing approaches in a safe environment,” says Hofmann. “Results like these will help us develop a better understanding on how to validate and debug reinforcement learning systems–a crucial step on the path towards real-world applications.”

Baker says there’s also a hopeful takeaway from the surprises in the hide-and-seek experiment. “If you put these agents into a rich enough environment they will find strategies that we never knew were possible,” he says. “Maybe they can solve problems that we can’t imagine solutions to.” Continue reading →

Posted in Human Robots