PNIPU the Robot for replacement of idle lamps the Robot for replacement of idle lamps

Main article: Industrial robots

2020: Development of the robot for replacement of idle lamps

Developers of the Perm Polytechnic University create a management system for the robot who will replace idle lamps. The artificial intelligence will increase quality of their service and will allow to avoid traumatizing people: these works carry out at height in adverse weather conditions. PNIPU reported about it on August 26, 2020. Algorithms of machine learning and neural networks will help to be guided with space to the robot.

Lamps on streets change at the height from 8 to 15 m. At the same time works carry out in the presence of rainfall, wind or low temperatures. Because of replacement of the lamp the transport network becomes more loaded. Specialists cannot carry out complete diagnostics of faults as they work in an uncomfortable position and are limited in time and tools. One robotic complex will allow to replace several crews which service lighting, – one of developers, the undergraduate of electrotechnical faculty of the Perm Polytechnic University Pavel Slivnitsin told.

That "teach" to be guided a robocomplex in the city and to interact with objects of the real world, scientists developed for it an operation algorithm. The artificial intelligence will recognize a lamp on the image which is transferred by the video camera of the robot, and directs his manipulator for replacement of the lamp. Algorithms and neural networks allow the robot to solve these problems in the conditions of different visibility, on different backgrounds and in a situation when an object is covered with something. The robocomplex is delivered to a lamp on the car which also the artificial intelligence can manage. That the robot could perform work, the lamp should have a universal connector. Fastening will allow to remove quickly it and to set, and replacement of the lamp will become the same transaction. The robot will execute it in the automatic mode.

In modern conditions when intelligent devices "go outside" from industrial workshops, it is important to them to be able to collect and process independently information, developers tell. The arrangement of objects in the real world can change therefore robots should not only work according to the instruction, but also be able to work in emergency situations.

Industrial robots on production are guided in space using the predeterminated cards or radio tags. But in open space, on the street, these technologies become expensive. It is impossible to equip with tags all objects and to support maps of the external environment in current status. Therefore we suggest to use neural networks and methods of machine learning that the robot distinguished difficult objects in real time. He classifies lamps by a photo or video and on the basis of it makes decisions, – the developer explained.

Scientists trained artificial intelligence at the separate server. The Perm developers provided training of neural network using 250 photos of lamps in "acquaintances" for the robot foreshortenings. The selection can be filled up further with new photos which are collected by a robocomplex. The device works effectively even at computing resources with a small power, emphasized in the Perm Polytechnic University.

Robotics