Skip to content

A Big Step for Microbatteries using Machine Learning

Research, In the News


Data-driven design of miniature batteries is key to maximizing their performance. Miniature lithium-ion batteries (microbatteries) have been promising candidates for powering off-grid operations of internet-of-things (IoT) devices. To further develop IoT devices, the power and energy densities of microbatteries must be increased. Considering the small size of microbatteries, three-dimensional (3D) design is one key approach to increasing their power and energy densities. However, designing such architecture manually is quite challenging even for experts. To overcome this challenge, The research team proposes an efficient architectural-design optimization method that consists of a geometry generator based on Monte Carlo tree search and prediction models that utilize machine learning. The proposed method completes geometry optimizations over 5.5 times more efficiently than the method based on a randomized algorithm and successfully designs state-of-the-art 3D battery architectures. The proposed approach is expected to be useful in designing microbatteries for various IoT electronics.

Link: Data-driven optimization of 3D battery design