Evolving Afferent Architectures: Biologically-inspired Models for Damage-Avoidance Learning
This addresses damage-avoidance learning for applications such as biomechanical systems, though it appears incremental as it builds on existing evolutionary and reinforcement learning methods.
The paper tackles the problem of damage-avoidance learning in long-horizon tasks like biomechanical digital twins by introducing Afferent Learning, which uses evolutionary optimization to discover sensing architectures that produce adaptive risk signals; this approach achieved a 23% reduction in high-risk actions compared to baselines.
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.