A Communication Consistent Approach to Signal Temporal Logic Task Decomposition in Multi-Agent Systems

TL;DR

This paper introduces a method for decomposing global Signal Temporal Logic tasks in multi-agent systems with limited communication, ensuring task consistency through a convex optimization-based approach.

Contribution

It proposes a novel task decomposition mechanism that aligns task dependencies with communication constraints, using decentralized convex optimization.

Findings

The method guarantees task satisfiability after decomposition under certain conditions.

It models task dependencies and communication limitations as graphs and optimizes task distribution.

The approach is applicable to systems with bounded polytope predicate functions.

Abstract

We consider the problem of decomposing a global task assigned to a multi-agent system, expressed as a formula within a fragment of Signal Temporal Logic (STL), under range-limited communication. Given a global task expressed as a conjunction of local tasks defined over the individual and relative states of agents in the system, we propose representing task dependencies among agents as edges of a suitably defined task graph. At the same time, range-limited communication naturally induces the definition of a communication graph that defines which agents have access to each other's states. Within these settings, inconsistencies arise when a task dependency between a pair of agents is not supported by a corresponding communication link due to the limited communication range. As a result, state feedback control laws previously derived to achieve the tasks' satisfaction can not be leveraged. We propose a task decomposition mechanism to distribute tasks assigned to pairs of non-communicating agents in the system as conjunctions of tasks defined over the relative states of communicating agents, thus enforcing consistency between task and communication graphs. Assuming the super-level sets of the predicate functions composing the STL tasks are bounded polytopes, our task decomposition mechanism can be cast as a parameter optimization problem and solved via state-of-the-art decentralized convex optimization algorithms. To guarantee the soundness of our approach, we present various conditions under which the tasks defined in the applied STL fragment are unsatisfiable, and we show sufficient conditions such that our decomposition approach yields satisfiable global tasks after decomposition.