One of the distinctive properties of living organisms are the programs of behavior inherent in us by nature, which we call instincts. And not the least of them, which allowed a person to ascend to the top of the evolutionary ladder, is the instinct of self-preservation. Thanks to him, we are able to protect ourselves from possible danger: be it caution when crossing the carriageway or refusal from night walks in unfavorable areas. But very soon, robots can also learn to protect themselves from danger.
Robotic mechanisms operate according to the program laid down in them and cannot deviate from it a single step, even if the robot rushes towards the abyss at full speed. DARPA specialists have developed a special algorithm that will allow self-driving mechanisms to predict a possible danger and thereby save themselves from damage or even destruction.
The DARPA algorithm in action. The flying drone calculates the flight path. Red lines indicate hazardous and potentially hazardous routes
During the tests of the self-preservation algorithm, DARPA indicated to the drone the starting and ending points of the route, which passed through the forest belt and was 1 kilometer. The results of several launches with different starting and ending points of the route were positive.
It's worth mentioning that DARPA's development is far from unique. Microsoft's research department has been working on such a system for several years. Tests are planned in the near future, but Microsoft's behavior algorithm looks more interesting. The functionality of the system includes not only the ability to analyze the situation, but also to remember the information received in order to avoid unpleasant situations in the future. Moreover, robots will be able to share maps of the area with each other indicating all dangerous routes.
Despite the fact that this algorithm brings robotic mechanisms even closer to the semblance of what they draw fantastic films about the future for us, it is too early to sound the alarm, because this is just a set of commands, in essence not much different from ordinary programs laid in the memory of machines … Until the machines gain a sense of identity, there is no need to worry about a bleak future. Although who knows what algorithms are being developed now within the framework of secret projects….
VLADIMIR KUZNETSOV
Promotional video: