Legged robots, particularly quadrupeds, have emerged as an essential research focus for traversing unstructured environments where wheeled systems are limited. This paper presents a novel Multi-Level Cognitive Control Architecture (MLCCA) for a quadruped robot, inspired by biological hierarchical intelligence. The primary novelty of this work lies in the integration of three bio-inspired cognitive layers—a fast reactive layer (Microscopic Brain), a mid-level adaptive layer (Mesoscopic Brain), and a high-level strategic planner (Macroscopic Brain)—into a unified, mathematically formalized control framework implemented entirely in MATLAB without requiring deep learning or extensive hardware. Unlike prior approaches that address either reflexive or deliberative control in isolation, the proposed architecture simultaneously handles immediate obstacle avoidance, terrain-adaptive gait modulation, and long-term path planning within a single coherent system. The robot dynamically alters its gait parameters based on local terrain gradients and obstacle height using height interpolation and cognitive decision rules. The simulation integrates procedural terrain generation and wireframe-based gait control visualization. Results demonstrate successful traversal of unpredictable terrain with intelligent gait adaptation and obstacle avoidance, achieving a mean body tilt of only 0.075rad and a path efficiency ratio of 1.18. This framework provides a scalable foundation for cognitive robotics, integrating perception, reflex, and long-term strategy into a unified control system.
N et al. (2026) studied this question.