Evolving Afferent Architectures: Biologically-inspired Models for Damage-Avoidance Learning

TL;DR

This paper presents a biologically-inspired framework called Afferent Learning that evolves sensing architectures to improve damage-avoidance policies in digital twins, demonstrating higher efficiency and age-robustness over traditional methods.

Contribution

It introduces a novel two-level evolutionary and reinforcement learning framework for designing afferent sensing architectures that enhance damage-avoidance learning in complex systems.

Findings

Evolved architectures outperform hand-designed baselines in efficiency and robustness.

CAT signals, evolution, and predictive discrepancy are validated as essential components.

Policies exhibit age-dependent behavioral adaptation with a 23% reduction in high-risk actions.

Abstract

We introduce Afferent Learning, a framework that produces Computational Afferent Traces (CATs) as adaptive, internal risk signals for damage-avoidance learning. Inspired by biological systems, the framework uses a two-level architecture: evolutionary optimization (outer loop) discovers afferent sensing architectures that enable effective policy learning, while reinforcement learning (inner loop) trains damage-avoidance policies using these signals. This formalizes afferent sensing as providing an inductive bias for efficient learning: architectures are selected based on their ability to enable effective learning (rather than directly minimizing damage). We provide theoretical convergence guarantees under smoothness and bounded-noise assumptions. We illustrate the general approach in the challenging context of biomechanical digital twins operating over long time horizons (multiple decades of the life-course). Here, we find that CAT-based evolved architectures achieve significantly higher efficiency and better age-robustness than hand-designed baselines, enabling policies that exhibit age-dependent behavioral adaptation (23% reduction in high-risk actions). Ablation studies validate CAT signals, evolution, and predictive discrepancy as essential. We release code and data for reproducibility.