Distributed Monitoring of Robot Swarms with Swarm Signal Temporal Logic

TL;DR

This paper introduces a distributed framework for monitoring robot swarms' behavior using Swarm Signal Temporal Logic, enabling agents to verify swarm properties with confidence through consensus algorithms and Kalman filtering.

Contribution

It presents a novel distributed monitoring approach for robot swarms that employs generalized moments, Kalman filters, and formal logic rules for real-time verification.

Findings

Effective estimation of swarm features using GMCA and Kalman filter

Bounded estimation error independent of agent motion

Successful application to a supply transportation swarm scenario

Abstract

In this paper, we develop a distributed monitoring framework for robot swarms so that the agents can monitor whether the executions of robot swarms satisfy Swarm Signal Temporal Logic (SwarmSTL) formulas. We define generalized moments (GMs) to represent swarm features. A dynamic generalized moments consensus algorithm (GMCA) with Kalman filter (KF) is proposed so that each agent can estimate the GMs. Also, we obtain an upper bound for the error between an agent's estimate and the actual GMs. This bound is independent of the motion of the agents. We also propose rules for monitoring SwarmSTL temporal and logical operators. As a result, the agents can monitor whether the swarm satisfies SwarmSTL formulas with a certain confidence level using these rules and the bound of the estimation error. The distributed monitoring framework is applied to a swarm transporting supplies example, where we also show the efficacy of the Kalman filter in the dynamic generalized moments consensus process.