Perlin Noise for Exploration in Reinforcement Learning
Keywords: Reinforcement Learning, Exploration Strategies, Perlin Noise, Policy Optimization, Structured Exploration
TL;DR: We propose a novel exploration strategy for Reinforcement Learning using Perlin Noise.
Abstract: Reinforcement Learning (RL) enables agents to solve tasks by autonomously acquiring policies by interacting with the environment receiving sparse or noisy feedback in the form of a reward. However, achieving successful optimization in RL requires efficient exploration, which remains a significant challenge, particularly in continuous action spaces. Existing exploration techniques often exhibit limited state-space reach and fail to overcome local optima, resulting in suboptimal policies. Additionally, these techniques can cause erratic movements, posing risks when applied to real-world robots.
In this work, we introduce a novel exploration strategy leveraging Perlin Noise, a gradient noise function that generates smooth, continuous disturbances, thus enhancing the agent's performance by promoting structured exploration and fluid motions. We quantitatively demonstrate the benefits of our approach compared to state-of-the-art methods, showing that it outperforms both unstructured and structured techniques in thorough experimental evaluations.
Primary Area: reinforcement learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 8188
Loading