ABSTRACT
This paper proposes a hierarchical reinforcement learning framework for decoupled pitch and azimuth control in multi-axis pointing systems, addressing challenges in traditional control approaches and leveraging temporal abstraction for improved control.
PAPER · PDF
Loading PDF...
Key findings
Proposes a novel hierarchical reinforcement learning framework termed Hierarchical Decoupled Controller (HDC).
Introduces a decoupled learning objective to address the coupling between pitch and azimuth dynamics.
Demonstrates superior tracking accuracy and robustness compared to conventional PID controllers and flat RL baselines in simulated pointing tasks.
Limitations & open questions
The application of HRL to continuous control problems, particularly multi-axis pointing systems, remains underexplored.