Living organisms have evolved over hundreds of millions of years to go the way of the first amphibians to mammals with large and complex brain. This way, only in a much shorter time frame, repeat, and robots. However, the contemporary scenario of development of robots is a significant shortcoming - the inability to respond to a change in their design, because the addition of any element or sensor makes substantial changes in the control unit software. This feature leads to higher and more time to develop autonomous systems, where each of the versions of software is developed and a new version of software. However, researchers from the Robert Gordon University, Aberdeen, United Kingdom, have created a robot that can respond to changes in its own design. In this case, the main challenge facing the engineers was the need to develop a flexible software that can learn over time, but most importantly pereuchivatsya if necessary. Based researchers have already well known to the public at large neural networks, which operate under the same algorithm with human brain cells - depending on the successful or unsuccessful outcome of certain actions are set, or vice versa, the rush of communication between certain nodes. designed by members of Robert Gordon University`s robot is a very simple mechanism, equipped for the start of only two front limbs, and a set of sensors, which controls the neural network. Before the device is quite a major task - to pass the largest possible distance for 1000 seconds. After many trial and error motion of the robot is trained, and that time is steadily taking place some distance - is likely to develop high speed will not allow construction of the device itself. At this stage of the procedure for teaching the robot deviates from the standard models of neural networks - software is the best version of the movement and uses it in the future. But the fun has yet to come - when changing the design of the two front limbs have found the mechanism of movement is not applicable in specific circumstances. In this case, traditional software is powerless before the new problem, but engineers developed an algorithm for incremental evolution (incremental evolutionary algorithm - IEA) will recognize that the design of the device has changed, and is required to conduct the process of retraining. But no less interesting that the IEA is able to not only respond to the changing of one element, but can detect the emergence of additional bodies, including the additional sensors. In this case also includes a re-stage neural network - to add two extra hind limbs developed by members of the University of Robert Gordon the robot drastically changes its mechanism of movement, adapting a new environment. Thus, the researchers managed to develop a simple algorithm that can react to changes in vehicle design and build a new model of behavior. Similarly, progress and evolution of living organisms, but most importantly, the development should form the basis of software to manage more complex systems. This IEA will help a new generation of robots is not only responsive to external stimuli, but also recognize the changes as positive or negative, of its own design.