Authors: Jonas Eschmann, Dario Albani, Giuseppe Loianno

Published on: June 06, 2023

Impact Score: 7.8

Arxiv code: Arxiv:2306.03530

Summary

• What is new: Presentation of RLtools, a pure C++ library, marking the introduction of Tiny Reinforcement Learning (TinyRL) by enabling deep reinforcement learning directly on microcontrollers.
• Why this is important: Deep Reinforcement Learning suffers from long training times and a lack of real-time applicability on embedded devices due to non-portable libraries.
• What the research proposes: RLtools, a dependency-free, header-only library, optimized for a wide range of platforms, significantly reducing training times and enhancing applicability in real-world devices.
• Results: RLtools achieved up to 76 times faster problem-solving capabilities compared to other RL frameworks and showcased the fastest inference on various microcontrollers.

Technical Details

Technological frameworks used: RLtools, a pure C++ library

Models used: Deep Reinforcement Learning algorithms

Data used: Various RL problem environments and microcontroller benchmarks

Potential Impact

Companies in robotics, IoT, and wearable technologies; markets related to smart devices, healthcare monitoring, and edge computing

Want to implement this idea in a business?

We have generated a startup concept here: MicroRL.