Researchers from the University of Lausanne developed an algorithm that allows employees to identify and isolate robots that can violate the consistency and efficiency of the entire swarm. The algorithm is based on the principle of most votes and uses local information about the behavior of neighboring robots.
Robots-employees are groups of autonomous robots that can perform various tasks in joint mode, such as studying the territory, search and rescue, or construction. However, these robots are also subject to possible malfunctions, errors, or malicious interventions that can cause some robots to act against the interests of the entire swarm. These robots are referred to as Byzant’s robots.
Byzant’s robots can create problems for the operation of the swarm, as they can transmit false or conflicting information to other robots or act against the rules. For example, a Byzant’s robot may indicate an incorrect direction of movement or refuse to follow the leader. This can lead to loss of synchronization, disunity, or even destruction of the swarm.
To prevent such situations, the researchers proposed an algorithm that allows the identification and isolation of Byzant’s robots. The blockchain-based algorithm works as follows: each robot observes the behavior of its neighbors and compares it with the expected behavior. If a neighbor’s behavior deviates beyond a certain threshold, the robot considers it suspicious and votes for its isolation. If most robots in the vicinity vote to isolate the suspicious robot, it is disconnected from communication and interaction with the rest.
The researchers conducted experiments with both real and virtual robots and demonstrated that the algorithm successfully detects and neutralizes Byzant’s robots in different scenarios. They also proved that the algorithm is resilient to measurement and communication errors, as well as random changes in the behavior of normal robots.