Volume 17, Number 2
S-AI-Robotics : A Sparse Artificial Intelligence Architecture with Hormonal Orchestration, Parsimonious Control, and Symbolic Memory for Adaptive, Safe, and Explainable Embodied Robotic
Authors
Said Slaoui , Mohammed V University, Morocco
Abstract
Robotic systems increasingly operate in dynamic, uncertain, and resource-constrained environments, where safety, energy efficiency, and explainability are as critical as raw performance. While learningbased and monolithic control architectures have demonstrated impressive capabilities, they often rely on continuous activation, data-intensive training, and opaque decision processes, making them fragile, energy-demanding, and difficult to audit in safety-critical contexts. This paper introduces S-AIROBOTICS, a bio-inspired and modular robotic intelligence framework grounded in the principles of Sparse Artificial Intelligence (S-AI). The proposed architecture departs from always-on robotic control by enforcing context-aware parsimony, where specialized robotic agents are activated only when justified by a symbolic hormonal state reflecting urgency, stability, energy, and confidence. A Robo-MetaAgent orchestrates agent activation through constrained optimization and hysteresis-based dynamics, ensuring stable and frugal behavior selection under competing objectives. To regulate behavioral priorities, S-AIROBOTICS integrates an artificial hormonal signaling layer, inspired by neuroendocrine systems, which modulates agent thresholds through bounded emission, inhibition, diffusion, and decay mechanisms. In parallel, a symbolic and contextual memory subsystem stores behavioral engrams—linking hormonal context, activated agents, actions, and outcomes—enabling rapid recall, adaptation, and native explainability of robotic decisions. The framework is evaluated using SAI-UT+, a reproducible experimental testbench, across multi-scenario robotic tasks including navigation, obstacle avoidance, energy scarcity, sensor degradation, and emergency handling. Results demonstrate that S-AI-ROBOTICS achieves improved stability, reduced energy consumption, and enhanced explainability compared to classical control, behavior trees, and reinforcement learning baselines, while maintaining robust performance under uncertainty. By unifying hormonal regulation, sparse orchestration, and symbolic memory within an embodied intelligence framework, S-AI-ROBOTICS establishes a principled foundation for adaptive, safe, and explainable robotic systems.