Filtering Inconsistent Failure in Robot Collective Decision with Blockchain

Abstract

Robotic swarm is one of typical and significant distributed systems. Fault tolerance is an important and essential ability for such systems because individual robot may malfunction and even maliciously attack, thereby interfering with reaching consensus on a decision at the swarm level. One kind of fault which is difficult to deal with is called inconsistent failure (ie. Byzantine failure). It is shown as that a robot exhibits different behavior at different times to other members of the swarm. Some previous work have proved that its feasible to exclude the interference of consistent failures via blockchain technology. We expanded the work to solve the problem in situation with inconsistent failures. In this paper, we proposed a method to filter inconsistent failures with blockchain in a collective decision task of robot swarm. The basic idea is to use the blockchain to record the behaviors of the robots, and then automatically audit these behaviors through the rules in the smart contract to achieve the purpose of detecting the robot with inconsistent failures. The feasibility and effectiveness of our approach is validated by a series of experiments in a voting scenario.