The first decentralized algorithm that could be very beneficial for self-driving vehicles in terms of such vehicles being able to navigate each other without crashing into one another or resulting traffic jams have been developed by researchers in the at the Northwestern University of the United States.
It has been reported that the researchers have already put the algorithm to test through a simulation of 1,024 robots and on a swarm of 100 real robots under laboratory conditions. The researchers said that with the help of the new algorithm, the robots were seen to reliably, safely and efficiently get together at a point according to a predetermined pattern within a time of less than a minute.
This was achieved by the researchers by increasing the power of the near-sight of the robots.
“Each robot can only sense three or four of its closest neighbors,” said Michael Rubenstein, a professor in Computer Science and Mechanical Engineering in Northern University’s McCormick School of Engineering who led the study. “They can’t see across the whole swarm, which makes it easier to scale the system. The robots interact locally to make decisions without global information.”
“In a decentralized system, there is no leader telling all the other robots what to do. Each robot makes its own decisions. If one robot fails in a swarm, the swarm can still accomplish the task,” Rubenstein said.
In order for them to avoid collisions and deadlock, the robots needed to coordinate among themselves while moving. This was achieved by the algorithm by viewing the ground beneath the robots in the form of a grid. And every robot fitted with the algorithm was able to become aware of where it sat on the grid through the use of a technology that was similar to GPS. Further, the robots also made use of sensors to communicate with its neighbors before making a decision about where to move. That helped the robots to determine whether the nearby spaces within the grid are vacant or occupied.
“The robots refuse to move to a spot until that spot is free and until they know that no other robots are moving to that same spot,” Rubenstein said. “They are careful and reserve a space ahead of time.”
In a disorganized form, a group of 1000 robots are able to coordinate with each other in the above mentioned manner for ultimately get into a predetermined shape in less than a minute’s time.
The researchers have claimed that fleets of driverless cars and automated warehouses could easily make use of this technology based on the new algorithm.
“Large companies have warehouses with hundreds of robots doing tasks similar to what our robots do in the lab,” Rubenstein said. “They need to make sure their robots don’t collide but do move as quickly as possible to reach the spot where they eventually give an object to a human.”
(Adapted from XinhuaNet.com)