Attention improves multi-agent communication in low bandwidth settings
Attention-based Hierarchical Variational Information Bottleneck for Robust Multi-Agent Communication under Variable Bandwidth
Machine LearningArtificial Intelligence
Summary
Communicating clearly between multiple agents is hard when only part of a message gets through. The authors study how to make messages that stay useful even if interrupted early. They propose a new method called AH-VIB that uses attention and a special compression technique to build messages piece by piece. Tested in a task where agents explore an environment with limited sensors, AH-VIB keeps communication reliable even when bandwidth is very limited, doing better than existing methods. This helps agents share useful info even when their messages get cut off.
What this means in practice
- •For robotics engineers: Build robotic teams that keep communicating useful information even when wireless signals are weak or limited.
- •For network protocol developers: Design message protocols that maintain meaningful data exchange when network bandwidth varies or messages get truncated.
Authors
Lukas Koch Vindbjerg, Qi Zhang, Yury Brodskiy, Lukas Esterle
Abstract
Learning-based multi-agent communication under limited bandwidth does not only require deciding what to communicate, but also structuring messages so that partial transmissions remain useful. We study this problem under prefix truncation, where only the first part of each message is received. To address it, we propose \textbf{AH-VIB}, an attention-based autoregressive variational communication model that combines a variational information bottleneck (VIB) with sequential message generation and a hierarchical robustness loss. We evaluate AH-VIB on a custom cooperative object-inspection and occupancy-mapping task, where agents equipped with a limited field-of-view sensor coordinate to scan inspection objects in an occupancy-grid world, under variable and fixed bandwidth conditions, and compare it against MADDPG, CommNet, a flat VIB baseline, and an autoregressive MLP ablation. AH-VIB achieves competitive mean return while improving performance reliability under the most constrained bandwidth conditions. These results indicate that AH-VIB improves the reliability and graceful degradation of learned communication under bandwidth constraints.