Self-robots and human consciousness – just different things. The main purpose of the development of robots endowed with self-consciousness, is to build systems capable of self-identified their unit to ensure the best possible adaptation to the environment functioning even in case of unforeseen changes in their own structure or parameters of the medium.
Also important quality of self-conscious machines is the possibility of their own identity to work effectively in a group of robots.
A story about identity robots should start with the entry of a young graduate student at Cornell University (Cornell University) Viktor Zukav. Born in Ivan, the famous “city of brides”, he graduated from Ivan State Power Engineering University. After receiving his diploma in 2002 at the Department of the electric, Victor received a master’s degree in Electro mechanics. After the Institute young man managed to change a few jobs before discovered that, despite the need for electric machines, similar needs for specialists on them is not observed.
Tosca young professionals dispelled one of the popular science broadcasts Discovery Channel, where the course Lipson (Hood Lipson), PhD, of Israeli origin from the private University of Massachusetts Brenda’s, talked about his unusual CAD-program, which allowed for hand-drawn outline of a wire model to restore a complete complex three-dimensional shape. It was also reported and the technology developed by American auto build robots from standardized elements, and finally Lipson shared the results of evolutionary algorithms.
Victor, inspired by the ideas of an American scientist, wrote him a letter in which, in turn, conveyed his thoughts about robotics, and asked how he could take part in its research activities. Lipson responded to a letter from Ivan, electrician, advising him to send his application for a place in graduate school at Cornell University, where the course of that time began to teach. Advertising Zukav was approved, and in March 2003, Victor was invited to study at university. There he joined the staff of the laboratory calculated synthesis (Computational Synthesis Laboratory), Where, together with the course Lipson and his former student Josh Bonged (Josh Bonged) From the University of Vermont (UMP) engaged in research in the field of evolutionary robotics. And one of the most recent work was the creation of a robot having “inner consciousness”.
According to laboratory staff, and their moving on 4 limbs child is able to independently assess the damage caused to him and determine how to adapt to continue the job.
Take for example a dog that has suddenly had a broken one paw. It is unlikely that the dog will rise or fall as if rooted to the ground. Rather, it is using three healthy legs and flee to a safe place. But for the majority of autonomous robots foot injuries like death. Most existing robots operate in an environment does not change the environment, but if they are to be useful outside, they must learn to adapt to unexpected changes.
The new robot, nicknamed because of its shape “Sea Star” (Starfish), “uses the relationship” perception – setting in motion “to examine its own structure and then use this” Sam model for movement. And when the leg is removed, the robot adapts its model to create alternative “gait.” This concept may help in developing new and complex machines and shed light on “samomodelirovanie” animals “.
“Most robots have a fixed model – a program laid by engineers – explains Lipson. – We are the first to show how the model can appear inside the robot. This makes the machine adaptable to a new level, because before they can pose the problem without worrying about creating program for its implementation. We hope that this – out of robots in the next step of knowledge.
“If a machine breaks down, it tends to break down completely,” – says Josh Bonged. – The main achievement is that the robot can adapt and continue to operate despite the damage. “
“Starfish” independently – without the written algorithm by man – has learned to walk, let her “walk” and looks ridiculous. Moves the robot, rather, on all fours, with some strange jerks. But the creators of “Sea Star” is a little concerned. “He has not been set to raise its body off the ground and walk.” He had a job to move the plane in a straight line, and a robot that problem was solved quite efficiently, “- says Victor Zukav. “World’s first robot to a challenge not to put in it people algorithms, and self-creating these algorithms” – proudly adds Proceedings Lipson.
“Initially, when the robot is switched on, it does not associate itself as something separate. The man looks at him and sees his four limbs and a small trunk. But the robot does not know this: he will receive this information during operation, when it begins to examine himself and those around the world “- says Lipson.
Immediately after the inclusion of “consciousness” of the robot is almost a “clean sheet”. The machine is the problem – move in a straight line. In addition, the robot is known for a set of elements, of which he is (responsible for motor 8 moving parts – 2 in each limb). From this data set it in my head randomly builds a hundred different combinations of assumptions about how to manage with all these mechanisms. Then sends a signal to the servo drives – and with built-in sensor checks the result. Of the 100 models of “first generation” 90% drop out, and the other slightly modified, “cross” with each other and re-tested to be useful to achieve the task. According to Zukav, this process is similar to how a young child to comprehend the world, trying, for example, all objects of taste. That’s “Starfish” initially may not try to run over an experiment, which is not quite logical. For example, try to knock yourself on the ground. Then she realizes that if the goal – to move forward, then slapped her on the ground more harm, and more so do not be.
After 16 cycles, each of which is based on 200 generations of virtual models. “Sea Star” finally “understands” how it is arranged, and remembers its selection of a virtual model. “Based on this model, the robot tries to achieve this mode of operation in which to achieve its objective and at the same time not destroy itself – tells Victor Zukav. According to him, in his first experience with their” Sea Star “has developed a system of governance legs that moved from their seats in the right direction at the first attempt.
“The machine is not a single model of itself – it has at the same time, there are many options that compete with each other,” – says Lipson. As a result, the robot generates an awkward, but quite functional gait, and the most effective option is still the way in which the robot is based on the “belly” and tightens your body, is repelled “hind” legs, reloaded, rolls.
Once the robot has learned to move with four legs, the scientists simulate an accident: disable the robot, tearing him one leg and include again. At the first step of the machine understands that the old, adapted to four feet of the virtual model does not correspond to the real sensor readings, and then stops. Then the “Starfish” is embarking on another round of evolutionary theories and develops a new algorithm of movement – for the three legs. Even more bizarre in appearance, but from no less effective.
The path length of several meters taken from “Sea Star” more than a day. But these ten ugly “steps” of a robot – a huge breakthrough for the whole robot.
For the experiment, the robot was built on four tentacles (scale of 0.6 meters). On board located single-board microcomputer PC-104 Production Advanced Digital Logic (based Celeron 266 MHz), with the right amount of signal inputs and is equipped with wireless module. But scientists have conferred on the controller is only a function of direct control of a robot, but the whole computational load to accelerate the experiment was moved to the University-owned multi-processor cluster. Despite its high performance, the calculations took about thirty hours. Moreover, each of the nodes in the cluster was working on the whole task, and thus simultaneously received dozens of independent results. This was necessary for statistical analysis and compare the options. Viktor Zukav said that there are so many options incarnation of evolutionary algorithms, and the choice of a particular time did not have too much importance.
According to the Advisor to the Director Central Research Institute of Robotics and Technical Cybernetics, Alexander Zheljeznyakova, this development will be demanded immediately, for example, to create a robot working in extreme conditions, whether the volcano crater or the surface of another planet. “But the most important thing is not it, but that in practice proved very possibility of creating a” thinking “tools. It’s one thing to build adaptive systems in word, another – to make them work”, – says Alexander ironstone. Therefore, such devices are much more interesting and useful than all the humanoid robots, which are now off in Japan, I’m sure an expert.
The main objective of this study was to demonstrate the efficiency of the concept of self-discovery, which was brilliantly done: the robot was able not only to personally assess the situation by building your own virtual model, but continued to move after the removal of any of the tentacles. In addition, their experiment science team Lipson proved that for a comfortable existence in the outside world for future robots do not need to have a detailed model of the body.
Jumbo “Sea Star”, his ability to represent themselves and adapt to itself set him apart from among their own kind. “We do not think that it’s really self-consciousness, when the robot is thinking about himself that he thinks – thinks Lipson. – But I really think he is on his way to consciousness, like a cat – there’s something at this level .
Bonged says the robot, in a sense, “conscious” at a primitive level, because he “thinks about itself.” Researchers have yet recognize that the robot itself is simple and yet modest success, but the basic algorithm can already be applied in a much more complex machines, to enable them to adapt to changes in the environment and for themselves to repair, replace parts.
“Robots on other planets must be able to continue their mission without human intervention, if they are damaged and can not report a problem on Earth,” – notes Bonged, and it is difficult to disagree.