Yohan Choi, Minjoon Kim, Jinsung Kim, Youn-Hee Han,

Journal of Advanced Technology Research, Vol. 10, No. 1, pp. 16-25, Jun. 2025
10.11111/JATR.2025.10.1.016, Full Text:
Keywords: Quadrupedal Robots, Deep Reinforcement Learning, Perceptive Locomotion, autonomous navigation, Sim-to-Real Transfer
Abstract

This paper aims to analyze and synthesize the latest research trends in locomotion and navigation technologies for quadrupedal robots based on 20 key publications. It highlights how simulation based Deep Reinforcement Learning (DRL) has emerged as a core paradigm in robot control, moving beyond traditional model-based approaches. The study systematically organizes major technological achievements and developments, focusing on Sim-to-Real transfer, advancements in perception technologies (blind control vs. vision-based control), the evolution of learning paradigms (End-to-End, hierarchical learning, knowledge distillation), and real-time adaptation strategies. By comparing and analyzing the novel learning methods and MDP (state, action, reward) designs proposed in each paper, the study presents the current technological landscape and suggests future research directions in the field.

Statistics
Cite this article