CATS: Temporally Regularized Smooth Reinforcement Learning Control
What happened
SungJae Ahn and other authors propose the CATS method, which combines a temporal penalty with linear ramp-up to suppress action oscillation in deep reinforcement learning policies. The paper states that the temporal penalty can constrain the expected action differences between current states that share the same successor state, thereby producing a spatial smoothing effect. The authors validated in simulation and real environments that CATS significantly reduces action oscillation without degrading task performance, and that its computational overhead is very small. This work has been submitted to ICRA 2027.
AI 根据报道生成 · 16 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROCATS: Smooth Control of Deep Reinforcement Learning Policies via Temporal Regularization
The authors propose the CATS method, which combines a temporal penalty with linear ramp-up to suppress action oscillations in deep reinforcement learning policies. The paper proves that the temporal penalty can bound the expected action differences between current states that share the same successor state, thereby producing a spatial smoothing effect. In simulations and real-world environments, the method is verified to significantly reduce action oscillations without degrading task performance, with very low computational overhead. The work was submitted to ICRA 2027.