Saloni Jain

Huihua Zhao, Rafael Cathomen, Lionel Gulich et al.

Recent advances in reinforcement learning (RL) have enabled impressive humanoid behaviors in simulation, yet transferring these results to new robots remains challenging. In many real deployments, the primary bottleneck is no longer simulation throughput or algorithm design, but the absence of systematic infrastructure that links environment verification, training, evaluation, and deployment in a coherent loop. To address this gap, we present AGILE, an end-to-end workflow for humanoid RL that standardizes the policy-development lifecycle to mitigate common sim-to-real failure modes. AGILE comprises four stages: (1) interactive environment verification, (2) reproducible training, (3) unified evaluation, and (4) descriptor-driven deployment via robot/task configuration descriptors. For evaluation stage, AGILE supports both scenario-based tests and randomized rollouts under a shared suite of motion-quality diagnostics, enabling automated regression testing and principled robustness assessment. AGILE also incorporates a set of training stabilizations and algorithmic enhancements in training stage to improve optimization stability and sim-to-real transfer. With this pipeline in place, we validate AGILE across five representative humanoid skills spanning locomotion, recovery, motion imitation, and loco-manipulation on two hardware platforms (Unitree G1 and Booster T1), achieving consistent sim-to-real transfer. Overall, AGILE shows that a standardized, end-to-end workflow can substantially improve the reliability and reproducibility of humanoid RL development.

Siva Kailas, Shalin Jain, Harish Ravichandar

Graph Neural Networks (GNNs), developed by the graph learning community, have been adopted and shown to be highly effective in multi-robot and multi-agent learning. Inspired by this successful cross-pollination, we investigate and characterize the suitability of existing GNN explanation methods for explaining multi-agent coordination. We find that these methods have the potential to identify the most-influential communication channels that impact the team's behavior. Informed by our initial analyses, we propose an attention entropy regularization term that renders GAT-based policies more amenable to existing graph-based explainers. Intuitively, minimizing attention entropy incentivizes agents to limit their attention to the most influential or impactful agents, thereby easing the challenge faced by the explainer. We theoretically ground this intuition by showing that minimizing attention entropy increases the disparity between the explainer-generated subgraph and its complement. Evaluations across three tasks and three team sizes i) provides insights into the effectiveness of existing explainers, and ii) demonstrates that our proposed regularization consistently improves explanation quality without sacrificing task performance.

Kevin Fu, Shalin Anand Jain, Pierce Howell et al.

Recent advances have enabled heterogeneous multi-robot teams to learn complex and effective coordination skills. However, existing neural architectures that support heterogeneous teaming tend to force a trade-off between expressivity and efficiency. Shared-parameter designs prioritize sample efficiency by enabling a single network to be shared across all or a pre-specified subset of robots (via input augmentations), but tend to limit behavioral diversity. In contrast, recent designs employ a separate policy for each robot, enabling greater diversity and expressivity at the cost of efficiency and generalization. Our key insight is that such tradeoffs can be avoided by viewing these design choices as ends of a broad spectrum. Inspired by recent work in transfer and meta learning, and building on prior work in multi-robot task allocation, we propose Capability-Aware Shared Hypernetworks (CASH), a soft weight sharing architecture that uses hypernetworks to efficiently learn a flexible shared policy that dynamically adapts to each robot post-training. By explicitly encoding the impact of robot capabilities (e.g., speed and payload) on collective behavior, CASH enables zero-shot generalization to unseen robots or team compositions. Our experiments involve multiple heterogeneous tasks, three learning paradigms (imitation learning, value-based, and policy-gradient RL), and SOTA multi-robot simulation (JaxMARL) and hardware (Robotarium) platforms. Across all conditions, we find that CASH generates appropriately-diverse behaviors and consistently outperforms baseline architectures in terms of performance and sample efficiency during both training and zero-shot generalization, all with 60%-80% fewer learnable parameters.