Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Performance-Aware Self-Configurable Multi-Agent Networks: A Distributed Submodular Approach for Simultaneous Coordination and Network Design

Zirui Xu, Vasileios Tzoumas

TL;DR

This work presents the AlterNAting COordination and Network-Design Algorithm (Anaconda), a scalable algorithm that also enjoys near-optimality guarantees and is an anytime self-configurable algorithm that quantifies its suboptimality guarantee for any type of network, from fully disconnected to fully centralized, and that is one order faster in terms of decision speed.

Abstract

We introduce the first, to our knowledge, rigorous approach that enables multi-agent networks to self-configure their communication topology to balance the trade-off between scalability and optimality during multi-agent planning. We are motivated by the future of ubiquitous collaborative autonomy where numerous distributed agents will be coordinating via agent-to-agent communication to execute complex tasks such as traffic monitoring, event detection, and environmental exploration. But the explosion of information in such large-scale networks currently curtails their deployment due to impractical decision times induced by the computational and communication requirements of the existing near-optimal coordination algorithms. To overcome this challenge, we present the AlterNAting COordination and Network-Design Algorithm (Anaconda), a scalable algorithm that also enjoys near-optimality guarantees. Subject to the agents' bandwidth constraints, Anaconda enables the agents to optimize their local communication neighborhoods such that the action-coordination approximation performance of the network is maximized. Compared to the state of the art, Anaconda is an anytime self-configurable algorithm that quantifies its suboptimality guarantee for any type of network, from fully disconnected to fully centralized, and that, for sparse networks, is one order faster in terms of decision speed. To develop the algorithm, we quantify the suboptimality cost due to decentralization, i.e., due to communication-minimal distributed coordination. We also employ tools inspired by the literature on multi-armed bandits and submodular maximization subject to cardinality constraints. We demonstrate Anaconda in simulated scenarios of area monitoring and compare it with a state-of-the-art algorithm.

Performance-Aware Self-Configurable Multi-Agent Networks: A Distributed Submodular Approach for Simultaneous Coordination and Network Design

TL;DR

This work presents the AlterNAting COordination and Network-Design Algorithm (Anaconda), a scalable algorithm that also enjoys near-optimality guarantees and is an anytime self-configurable algorithm that quantifies its suboptimality guarantee for any type of network, from fully disconnected to fully centralized, and that is one order faster in terms of decision speed.

Abstract

We introduce the first, to our knowledge, rigorous approach that enables multi-agent networks to self-configure their communication topology to balance the trade-off between scalability and optimality during multi-agent planning. We are motivated by the future of ubiquitous collaborative autonomy where numerous distributed agents will be coordinating via agent-to-agent communication to execute complex tasks such as traffic monitoring, event detection, and environmental exploration. But the explosion of information in such large-scale networks currently curtails their deployment due to impractical decision times induced by the computational and communication requirements of the existing near-optimal coordination algorithms. To overcome this challenge, we present the AlterNAting COordination and Network-Design Algorithm (Anaconda), a scalable algorithm that also enjoys near-optimality guarantees. Subject to the agents' bandwidth constraints, Anaconda enables the agents to optimize their local communication neighborhoods such that the action-coordination approximation performance of the network is maximized. Compared to the state of the art, Anaconda is an anytime self-configurable algorithm that quantifies its suboptimality guarantee for any type of network, from fully disconnected to fully centralized, and that, for sparse networks, is one order faster in terms of decision speed. To develop the algorithm, we quantify the suboptimality cost due to decentralization, i.e., due to communication-minimal distributed coordination. We also employ tools inspired by the literature on multi-armed bandits and submodular maximization subject to cardinality constraints. We demonstrate Anaconda in simulated scenarios of area monitoring and compare it with a state-of-the-art algorithm.
Paper Structure (20 sections, 7 theorems, 20 equations, 3 figures, 3 algorithms)

This paper contains 20 sections, 7 theorems, 20 equations, 3 figures, 3 algorithms.

Table of Contents

  1. Introduction
  2. Anytime Self-Configuration
  3. Decision Speed
  4. Approximation Performance
  5. Distributed Simultaneous Coordination and Network Design
  6. Alternating Coordination and Network-Design Algorithm ( Anaconda)
  7. Multi-Armed Bandit Problem
  8. Action Coordination
  9. Neighbor Selection
  10. Approximation Guarantee
  11. Runtime Analysis
  12. Numerical Evaluation in Sensor Scheduling for Area Monitoring
  13. Anaconda with different neighborhood sizes
  14. DFS-SG konda2022execution
  15. Conclusion
  16. ...and 5 more sections

Key Result

Proposition 1

The agents select actions via ActionCoordination such that: where $\tilde{O}(\cdot)$ hides $\log$ terms.

Figures (3)

  • Figure 1: Overview of AlterNAting COordination and Network-Design Algorithm ( Anaconda). Starting from an unspecified communication network, and subject to the agents' communication bandwidth and connectivity constraints, Anaconda enables the agents to optimize their local communication neighborhoods such that the action-coordination approximation performance of the whole network is maximized. To this end, Anaconda employs two subroutines, ActionCoordination and NeighborSelection, that alternate optimization. In more detail, given the selected neighborhoods $\{{\@fontswitch\mathcal{N}}_i\}_{i\in{\@fontswitch\mathcal{N}}}$ by NeighborSelection, ActionCoordination instructs the agents to select actions to jointly maximize \ref{['eq:intro']}. But ActionCoordination incurs a suboptimality cost $C(\{{\@fontswitch\mathcal{N}}_i\}_{i\in{\@fontswitch\mathcal{N}}})$ due to requiring the agents to coordinate exchanging local information only, prohibiting also multi-hop communication, in favor of decision speed. For this reason, given the agents' bandwidth and connectivity constraints, and the previously selected actions by ActionCoordination, NeighborSelection instructs each agent $i$ to design its neighborhood ${\@fontswitch\mathcal{N}}_i$ to optimize $C(\{{\@fontswitch\mathcal{N}}_i\}_{i\in{\@fontswitch\mathcal{N}}})$ and, thus, maximize the approximation performance of ActionCoordination in the subsequent iteration.
  • Figure 2: Area Monitoring with Multiple Cameras: Anaconda vs. DFS-SG. The cameras select the locations of their FOVs either per Anaconda with different maximum neighborhood sizes ranging among $\{0,1,3,5\}$, or per the DFS-SG. (a)-(c) are averaged over $30$ Monte-Carlo trials. From (a) to (b) to (c), the time $\tau_c$ of communicating an action decreases compared to the time $\tau_f$ of completing a function evaluation, with $\tau_c/\tau_f=\{5, 1, 0.2\}$.
  • Figure : Area Monitoring with Multiple Cameras: Anaconda vs. DFS-SG. The cameras select the locations of their FOVs either per Anaconda with different maximum neighborhood sizes ranging among $\{0,1,3,5\}$, or per the DFS-SG. (a)-(c) are averaged over $30$ Monte-Carlo trials. From (a) to (b) to (c), the time $\tau_c$ of communicating an action decreases compared to the time $\tau_f$ of completing a function evaluation, with $\tau_c/\tau_f=\{5, 1, 0.2\}$.

Theorems & Definitions (14)

  • Definition 1: Normalized and Non-Decreasing Submodular Set Function fisher1978analysis
  • Definition 2: 2nd-order Submodular Set Function crama1989characterizationfoldes2005submodularity
  • Remark 1: Decision speed vs. Optimality
  • Proposition 1: Approximation Performance
  • Definition 3: Mutual Information between an Agent and Its Neighbors
  • Lemma 1: Monotonicity and Submodularity of $I_{f,\,t}$
  • Theorem 1: Approximation Performance
  • Remark 2: On the Tightness of Approximation Bounds
  • Proposition 2: Computational Complexity
  • Proposition 3: Communication Complexity
  • ...and 4 more