Clone Swarms: Learning to Predict and Control Multi-Robot Systems by Imitation

Abstract

In this paper, we propose SwarmNet -a neural network architecture that can learn to predict and imitate the behavior of an observed swarm of agents in a centralized manner. Tested on artificially generated swarm motion data, the network achieves high levels of prediction accuracy and imitation authenticity. We compare our model to previous approaches for modelling interaction systems and show how modifying components of other models gradually approaches the performance of ours. Finally, we also discuss an extension of SwarmNet that can deal with nondeterministic, noisy, and uncertain environments, as often found in robotics applications.

Local Repair of Neural Networks Using Optimization

“…3. Also, let E(W l , b l ) be the loss function defined in (2). Repair the weight and bias vectors W l and b l of the l th layer in NN o (as shown in Fig.…”

End-to-end decentralized formation control using a graph neural network-based learning method

“…Breakthroughs have been made in computational methods that address long-standing challenges in multi-robot learning, such as non-stationarity (Foerster et al, 2017;Lowe et al, 2017;Foerster et al, 2018), learning to communicate (Foerster et al, 2016;Sukhbaatar and Fergus, 2016;Jiang and Lu, 2018), and scalability (Gupta et al, 2017). Various multi-robot control problems, such as path planning (Wang et al, 2021;Blumenkamp et al, 2022) and coordinated control (Zhou et al, 2019;Agarwal et al, 2020;Tolstaya et al, 2020a;Tolstaya et al, 2020b;Jiang and Guo, 2020;Kabore and Güler, 2021;Yan et al, 2022), have been tackled using learning-based methods. Despite the remarkable progress that has been made in multi-robot learning, the best approaches to architecture design and learning for scalable computational models that accommodate emerging information structures remain an open question.…”

“…A GNN can be trained to capture task-relevant information to be propagated and shared within the robot team via local inter-robot communication. GNNs have become an appealing framework for modeling of distributed robot networks (Agarwal et al, 2020;Zhou et al, 2019;Tolstaya et al, 2020a;Wang and Gombolay, 2020;Blumenkamp et al, 2022) due to their scalability and permutation-invariance (Gama et al, 2020).…”