ModGNN: Expert Policy Approximation in Multi-Agent Systems with a Modular Graph Neural Network Architecture

Ryan, Kortvelesy,; Prorok, Amanda

doi:10.1109/icra48506.2021.9561386

Cited by 20 publications

(10 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This was concurrently developed by a line of work that builds on Graph Neural Networks (GNNs), which are permutation equivariant by design [127][128][129]. GNNs have since then shown promising results in learning explicit communication strategies that enable complex multi-agent coordination [130][131][132][133].…”

Section: Learning Communication Behaviorsmentioning

confidence: 99%

A Critical Review of Communications in Multi-Robot Systems

Gielis¹,

Singh²,

Prorok³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Purpose of Review. This review summarizes the broad roles that communication formats and technologies have played in enabling multi-robot systems. We approach this field from two perspectives: of robotic applications that need communication capabilities in order to accomplish tasks, and of networking technologies that have enabled newer and more advanced multi-robot systems. Recent Findings. Through this review, we identify a dearth of work that holistically tackles the problem of co-design and co-optimization of robots and the networks they employ. We also highlight the role that data-driven and machinelearning approaches play in evolving communication pipelines for multi-robot systems. In particular, we refer to recent work that diverges from hand-designed communication patterns, and also discuss the 'sim-to-real' gap in this context. Summary. We present a critical view of the way robotic algorithms and their networking systems have evolved, and make the case for a more synergistic approach. Finally, we also identify four broad Open Problems for research and development, while offering a data-driven perspective for solving some of them.

show abstract

Section: Learning Communication Behaviorsmentioning

confidence: 99%

A Critical Review of Communications in Multi-Robot Systems

Gielis¹,

Singh²,

Prorok³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This has thus provided alternatives for the aforementioned challenges [5], [6], [7], [8]. Graph Neural Networks (GNNs), in particular, demonstrate remarkable performance and generalize well to large-scale robotic teams for various tasks such as flocking, navigation, and control [9], [6], [10], [11], [12], [13]. In such multi-robot systems, GNNs learn inter-robot communication strategies using latent messages.…”

Section: Introductionmentioning

confidence: 99%

“…Even though GNNs have an inherently decentralizable mathematical formulation, previous work on GNN-based multi-robot policies was conducted exclusively in centralized simulations using synchronous communication [10], [11], [9]. For practical reasons, decentralized execution is often unavoidable in the real-world, but it is currently unknown whether this contributes to a shift of domains, and how resulting policies are affected.…”

Section: Introductionmentioning

confidence: 99%

A Framework for Real-World Multi-Robot Systems Running Decentralized GNN-Based Policies

Blumenkamp¹,

Morad²,

Gielis³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) are a paradigmshifting neural architecture to facilitate the learning of complex multi-agent behaviors. Recent work has demonstrated remarkable performance in tasks such as flocking, multi-agent path planning and cooperative coverage. However, the policies derived through GNN-based learning schemes have not yet been deployed to the real-world on physical multi-robot systems. In this work, we present the design of a system that allows for fully decentralized execution of GNN-based policies. We create a framework based on ROS2 and elaborate its details in this paper. We demonstrate our framework on a case-study that requires tight coordination between robots, and present firstof-a-kind results that show successful real-world deployment of GNN-based policies on a decentralized multi-robot system relying on Adhoc communication. A video demonstration of this case-study can be found online 1 .

show abstract

“…Hand-engineered coordination strategies often fail to deliver the desired performance, and despite ongoing progress in this domain, they still require substantial design effort. Recent work has shown the promise of Graph Neural Networks (GNNs) to learn explicit communication strategies that enable complex multi-agent coordination[29,30,3,4]. In the context of multi-robot systems, individual robots are modeled as nodes, the communication links between them as edges, and the internal state of each robot as graph signals.…”

mentioning

confidence: 99%

“…Since compression is performed on local networks (with parameters that can be shared across the entire graph), GNNs are able to compress previously unseen global states. In the process of learning how to compress the global state, GNNs also learn which elements of the signal are the most important, and discard the irrelevant information[29]. This produces a non-injective mapping from global states to latent states, where similar global states 'overlap', further improving generalization.…”

mentioning

confidence: 99%

The Holy Grail of Multi-Robot Planning: Learning to Generate Online-Scalable Solutions from Offline-Optimal Experts

Prorok¹,

Blumenkamp²,

Li³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Many multi-robot planning problems are burdened by the curse of dimensionality, which compounds the difficulty of applying solutions to large-scale problem instances. The use of learning-based methods in multi-robot planning holds great promise as it enables us to offload the online computational burden of expensive, yet optimal solvers, to an offline learning procedure. Simply put, the idea is to train a policy to copy an optimal pattern generated by a small-scale system, and then transfer that policy to much larger systems, in the hope that the learned strategy scales, while maintaining near-optimal performance. Yet, a number of issues impede us from leveraging this idea to its full potential. This blue-sky paper elaborates some of the key challenges that remain.

show abstract

ModGNN: Expert Policy Approximation in Multi-Agent Systems with a Modular Graph Neural Network Architecture

Cited by 20 publications

References 4 publications

A Critical Review of Communications in Multi-Robot Systems

A Critical Review of Communications in Multi-Robot Systems

A Framework for Real-World Multi-Robot Systems Running Decentralized GNN-Based Policies

The Holy Grail of Multi-Robot Planning: Learning to Generate Online-Scalable Solutions from Offline-Optimal Experts

Contact Info

Product

Resources

About