One of the many benefits of self-driving cars in the future is to ease traffic congestion and avoid traffic accidents. A team from Northwestern University recently proposed a novel control algorithm that claims to provide a traffic-free environment and demonstrated an algorithmic model through a group of micro-robots. These miniature robots can quickly assemble shapes in 60 seconds.
The team at Northwestern University described the new control software as the first decentralization algorithm without collisions and deadlocks. Decentralizing algorithms can enhance the reliability of the system by avoiding the spread of the risk of failure across the fleet without the need for a centralized large computing center or leading robot.
Michael Rubenstein, who led the study, said: “If the system is centralized, once a robot stops working, the whole system fails.” In a decentralized system, no leader tells all other robots what to do. Every robot makes its own decisions. Instantly if a robot goes wrong in a cluster, the cluster can still complete the task. “
The team’s algorithm treats the terrain as a grid, and each robot is equipped with GPS-like sensors that can always know where they are on the grid. Other sensors allow the robot to identify nearby robots and know whether space on the grid is available or occupied by neighbors.
“Each robot can only sense three or four of its immediate neighbors,” Rubenstein explains. They don’t see the entire group, which makes it easier to scale the system. Robots can interact locally to make decisions without global information. “
This decentralized algorithm moves faster to form the desired shape. In the tests, a team of 100 robots was able to complete a scheduled operation in 60 seconds, compared with the previous method, which took up to an hour. They also tested the method in simulations of more than 1,000 robots and safely and effectively moved virtual machines into formations.
The study, published in the journal IEEE Transactions on Robotics, can be found in the following dynamic chart.