Papers
7
Total Citations
120
H-Index
4
About
Tadashi Horiuchi is a leading researcher in reinforcement learning and autonomous robotics, with a career spanning foundational theory to modern deep-learning applications. His work centers on enabling robots to acquire complex behaviors through trial-and-error learning, bridging classical algorithms and cutting-edge neural architectures. Horiuchi’s most influential contribution is his pioneering application of Deep Q-Networks (DQN) to vision-based mobile robot control (78 citations), demonstrating how convolutional neural networks can learn action-value functions directly from camera input—a milestone that helped popularize deep reinforcement learning in robotics. He also introduced the Q-PSP Learning algorithm (18 citations), an exploitation-oriented approach that prioritizes reinforcing successful experiences over exhaustive exploration, offering a practical alternative for real-time systems. His incremental state-space construction method (10 citations), inspired by Piaget’s theory of contradiction, adaptively builds state representations during learning, reducing computational overhead. Beyond wheeled robots, Horiuchi has extended reinforcement learning to snake-like and four-legged locomotion, integrating central pattern generators with multi-objective optimization. His work in the RoboCup Small Size League further showcases his commitment to real-world multi-robot coordination. With over 120 total citations, Horiuchi’s research continues to shape how autonomous agents learn from interaction.
Research Focus
Key Achievements
Top Papers
- 1A study on vision-based mobile robot learning by deep Q-network78 citations · 2017
- 2
- 3
- 4
- 5
- 6
- 7CPG-BASED LOCOMOTION LEARNING OF FOUR-LEGGED ROBOT BY MULTI-OBJECTIVE GA2 citations · 2014