Shared Spatial Memory Through Predictive Coding

2025-11-06

AI Summary

In multi-agent systems, partial observability and limited communication bandwidth induce spatial memory inconsistency among agents. Method: This paper proposes a unified predictive coding framework that models cooperation as mutual information minimization, enabling efficient communication and joint uncertainty reduction. The approach integrates self-supervised motion prediction, an information bottleneck constraint, multi-agent predictive coding, and hierarchical reinforcement learning. Contribution/Results: It spontaneously emergent grid-cell-like spatial encoding and hippocampal-like social place representations, while supporting bandwidth-adaptive active exploration. Evaluated on the Memory-Maze benchmark, our method maintains task success rates of 64.4% under extreme bandwidth compression—from 128 to 4 bits per step—only marginally down from 73.5%, significantly outperforming the baseline (28.6%). This demonstrates both biological plausibility and engineering robustness in resource-constrained cooperative navigation.

Technology Category

• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Social networks and social learning
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Sharing and reconstructing a consistent spatial memory is a critical challenge in multi-agent systems, where partial observability and limited bandwidth often lead to catastrophic failures in coordination. We introduce a multi-agent predictive coding framework that formulate coordination as the minimization of mutual uncertainty among agents. Instantiated as an information bottleneck objective, it prompts agents to learn not only who and what to communicate but also when. At the foundation of this framework lies a grid-cell-like metric as internal spatial coding for self-localization, emerging spontaneously from self-supervised motion prediction. Building upon this internal spatial code, agents gradually develop a bandwidth-efficient communication mechanism and specialized neural populations that encode partners'locations: an artificial analogue of hippocampal social place cells (SPCs). These social representations are further enacted by a hierarchical reinforcement learning policy that actively explores to reduce joint uncertainty. On the Memory-Maze benchmark, our approach shows exceptional resilience to bandwidth constraints: success degrades gracefully from 73.5% to 64.4% as bandwidth shrinks from 128 to 4 bits/step, whereas a full-broadcast baseline collapses from 67.6% to 28.6%. Our findings establish a theoretically principled and biologically plausible basis for how complex social representations emerge from a unified predictive drive, leading to social collective intelligence.

Problem

• Shared spatial memory challenge in multi-agent partial observability systems
• Bandwidth-efficient communication for multi-agent coordination under constraints
• Developing biologically plausible social representations through predictive coding

Innovation

• Multi-agent predictive coding minimizes mutual uncertainty
• Grid-cell-like spatial coding emerges from motion prediction
• Hierarchical reinforcement learning reduces joint uncertainty