To explore our solar system will require a new breed of intelligent robots
Science at NASA
Direct link to this page: https://www.hayadan.org.il/robo_helpers.html
When he shouts "Bring it on, Rover!" The man throws a frisbee far to the edge of the yard. His dog just stands there and does nothing.
Oh, yes, he thinks, I forgot. "Rover! Turn clockwise until you're facing north-northwest, move forward ten-point-five meters then stop, lower your head fourteen inches then close your mouth on the frisbee" he instructs. "And then… go back!"
The dog skips to the edge of the yard, only to return with an empty mouth. The frisbee was two centimeters away from what the man said - but the dog was unable to make this small correction on its own. So the man gives the tedious instruction again, this time saying "ten-point-five-two".
Rover plays fetch the Frisbee. Lawrence Manning.
Playing like this is not fun at all!
But it could have been worse. What if you were an astronaut who had just landed on the surface of Mars and a rover was your robotic assistant? When you only have a few precious months to explore Mars, a rover would be a colossal waste of time. Without common sense and flexibility, a rover can even endanger you in the unforgiving environment of a distant planet. He may blindly follow instructions that would cause him to damage vital equipment, and he will ignore requests in the plain language of an astronaut in distress.
The Sojourner rover that explored the surface of Mars in 1997 acted much like our helpless dog. Teams of scientists here on Earth had to feed Sojourner precise step-by-step instructions for each task he performed. If the vehicle hit an obstacle, it simply stopped and waited. The scientists had to tell him exactly how to overcome the problem. It took days to do simple tasks.
But Sojourner nevertheless succeeded thanks to the ingenuity and patience of its operators. However, much more was possible. If we're serious about exploring the solar system, mission planners say, we must build smarter, more capable robots.
Robots with common sense
"Over the next decade," says NASA Ames roboticist Liam Pederson, "there probably won't be a human presence much beyond Earth's orbit. So if we want to explore places like Mars, we're going to have to send robots. No robots, no space exploration. Period."
"Transmitting detailed instructions to robots that are essentially stupid is very expensive and inefficient - especially when there is a lot to do," he adds. For example: robots patrolling Mars, perhaps as a prelude to manned exploration, will cover large areas, they will sample hundreds of rocks, drill holes in search of frozen water and take thousands of pictures. "If each of these operations will require several days and an army of control personnel... well, you can understand how the cost increases."
In 1997, the SUV Sojourner "sniffs" a sedimentary rock named Yogi.
The first men on Mars will be just as busy as the rovers before them. The astronauts will have to establish the first base on an extraterrestrial world and learn to survive in a place compared to which Antarctica seems pleasant. And while they are busy with this, they will collect thousands of measurements for the scientists in Israel.
"The astronaut's time will be more valuable than edible gold," says Pedersen. "They will need intelligent robotic assistants."
How smart? The same intelligence that we usually take for granted in animals will suffice, says Pedersen. Animals effortlessly distinguish between the objects in their environment based on input from their senses. They are able to detect threats, and they intuitively understand how objects move and behave. They can identify targets - such as a small, scurrying piece of food - and then plan and carry out all the actions to get it. And they know their energy, power, temperature and endurance limits and are careful not to exceed them.
Getting a robot to do all this is not simple.
Says Pedersen, "Try teaching a robot this simple thing: 'You can't turn a glass of water over because the water will spill out.'
The computer brains of ordinary robots work basically the same way as home computers: they execute a fixed program of "if-then" logic and calculations. The speed and accuracy of this approach make computers the best at limited and specialized tasks. But it also makes them inflexible. Amat is a normal robot with a situation outside of his programming and has no idea how to react.
It turns out that it is very difficult to duplicate the adaptability and original problem-solving ability of humans (and of many animals).
learning from experience
Nevertheless, a patchwork of approaches to more flexible computation have emerged. Among them are technologies such as probability theory, genetic programming, natural language recognition and neural networks. Each allows to add learning or flexibility to the robot.
For example, scientists at Carnegie Mellon University taught a robot to drive a car automatically for 98% of a trip across the United States - a project cleverly called "No Hands Across America". We initially trained the robot by having it drive a car and watch a human driving the car. The robot has learned to associate certain visual inputs with the correct steering responses.
The "brain" of this robot was a computer simulation of a neural network, basically mimicking the architecture of animal brains. Input signals are processed by networks of "nodes" (neurons) and "connections" (axons). Neural networks learn from experience and can associate general inputs with specific outputs: four legs + barking (the inputs) = dog, for example.
In the picture: a silly example of a neural network. Input signals enter from the left, pass through both processing layers, and then appear from the right as output signals. This architecture can perform incredibly sophisticated logic, especially when adding feedback loops.
A site dealing with neural networks
Pederson warns that our understanding of the inner workings of organic minds is not enough to mimic them accurately. "Although neural networks are in some ways similar to organic brains," he says, "they are still much less complex or capable."
Probability theory, and especially Bayesian statistics, offers a different route to machine learning, says Pedersen. It allows computers to operate not only in terms of black and white - true or false - but also in shades of gray. Machines that "think" using such statistical models learn well from new and unexpected experiences. ("In my opinion, this is the exciting part of robotics," notes Pederson. "Expect an explosion in the capabilities of the robots").
Another option is genetic programming (evolutionary computing), where computers "develop" their own software. "Mutations" of an original program are tested, and those that give better results are kept. Their code is mixed and mutated again - like sexual reproduction - to create the next "generation", and so on for hundreds or thousands of generations. This software "evolution" can produce very efficient problem-solving programs that are too complicated for the scientists themselves to understand.
Above: The human brain - we all have one, but its inner workings are mysterious. If researchers learn more about organic brains it may help them program smarter robots. Grey's Anatomy.
This approach and other innovative approaches to calculation form the basis for smarter and more autonomous robots. The scientists take from this toolbox to include in the robots the same abilities that we take for granted in relation to ourselves: understanding the meaning of spoken language, understanding all the small actions necessary to complete a task, navigating the terrain and avoiding dangers - the practical matters of an autonomous tour.
In search of R2-D2
Indeed there is progress. One robot prototype called Hyperion has demonstrated the ability to independently cross the Canadian Arctic. This robot, developed by researchers at Carnegie Mellon University's Robotics Institute, carefully navigates to avoid getting stuck in the shade, so its solar panels capture sunlight at all times. And he's smart enough to know when he's lost or in trouble.
Another experimental robot, called the "robotic assistant for activities outside the vehicle" is a real partner of the astronauts - wandering on wheels next to a person in a spacesuit. Scientists at the Johnson Space Center are using it to test advanced technologies such as natural language interaction and astronaut gesture recognition. Much of what they learn will help in the development of similar aids, not only for planetary surfaces but also for Earth orbit and deep space.
Pictured: NASA's "robot assistant for extra-vehicular activities" next to an astronaut in a space suit. The two are real partners on the tour.
Pedersen notes: "Here at Ames we are working on a vehicle called K9 that will be able to do many things on its own. It can look at rocks, take measurements and decide what is 'interesting.' Mars, 9 (also known as the Smart Landing Vehicle on Mars and the Mobile Laboratory).
Other experimental robots are pioneers in another field: living in a spaceship. The Personal Satellite Assistant (PSA), for example, is a small floating ball that can propel itself with propellers inside spacecraft corridors. The PSA was created by Yuri Gudiak and his colleagues at NASA Ames and looks strikingly similar to Luke Skywalker's robotic lightsaber training partner from Star Wars. That's no accident, says Gudiak, who invented the PSA after seeing the film.
PSA will be able to do many things: talk to astronauts who want information from the ship's main computer; monitor the air (like a canary in a coal mine) to detect concentrations of gases that have a possible risk, for example too much carbon dioxide; or simply enter situations that may involve too much danger or uncertainty for their human teammates. Such hi-tech assistants would be welcome on the International Space Station.
(Left) An artist's illustration of Robonaut working outside the spaceship. (Center) Yuri Gudiak of NASA Ames and his personal satellite assistant. (Right) The K9 smart vehicle in field tests at NASA Ames.
Other robots are best suited for work outside the spacecraft. Robonaut, for example, is under development at the Johnson Space Center. He has a basic form of a man - or actually half a man. His body ends at the waist. His arms and hands are designed to be highly dexterous and his head contains video cameras. Astronauts, safely inside their spacecraft, will be able to perform routine maintenance or important repairs on the exterior of the spacecraft using Robonaut as a remote-controlled messenger.
If robots are going to live in spaceships, Pedersen points out, then you have to think about the robots when designing the spaceships. "The need for this kind of planning at the system level - designing the robot and the spacecraft so that each fits the other - is often overlooked by non-experts," he says. The spacecraft must have facilities for charging and storing the robots, and the robot must be able to access the spacecraft's computers and handle any necessary equipment.
The International Space Station and its robotic arm, Canadarm2, are an example of a well-integrated system. The arm crawls on the outer side of the spaceship - flipping end over end like a caterpillar from one specially placed outpost to the next. A custom built cart can move the arm quickly from place to place when speed is important.
Canadarm2 is impressive, but like Sojourner on Mars it is neither smart nor independent. The arm only moves when it receives commands from a person.
Right: working together. Astronaut Jerry Ross hovers above Earth, attached to one end of Canadarm2.
The main reason for the "wisdom gap" between the robots in the scientists' laboratories (such as K9) and those that have flown in space is the lack of proven reliability. Pedersen explains: "The problem is that these advanced technologies do not have any flight history. Will they operate under the demanding conditions of space flight? The mission managers They are rightly conservative; they prefer to stick to well-proven solutions."
However, with time and with field tests, the best of these technologies will prove their character - or actually, their silicon. And that's a good thing, because future astronauts will want their silicon assistants.
Translated by Nahum Sherashevsky, professional translation and technical translation, NIS 45 per 250 words (in the Hebrew version). Dew. 02-6435139
