Data-Driven Reinforcement Learning for Optimal Motor Control in Washing Machines

2024 IEEE Conference on Artificial Intelligence (CAI), Singapore

Chanseok Kang Guntae Bae Daesung Kim Kyoungwoo Lee Dohyeon Son Chul Lee Jaeho Lee Jinwoo Lee Jae Woong Yun

LG Electronics AI Lab

IEEE DOI

TL;DR

Delayed Online Update combines real washing-machine operation data with continual offline reinforcement learning to improve load balancing during dehydration without hand-coding every laundry scenario.

Delayed online update framework for continual offline reinforcement learning.

Delayed Online Update (DOU) workflow for post-deployment improvement.

Overview

Problem

Modern washing machines must keep loads balanced across many operating conditions. Manual trial-and-error tuning can limit how quickly control performance improves after deployment.

Approach

The paper introduces a continual offline reinforcement learning workflow for washer motor control, using logged transition data and delayed online updates to reduce distribution-shift risk.

Outcome

Experiments report improved balance-maintenance behavior during the dehydration cycle across tasks with different laundry conditions.

Method

1. Collect transition data from real washing-machine operation.
2. Train an offline RL policy for dehydration-cycle motor control.
3. Accumulate new online interaction data over a delay window.
4. Update the policy from the expanded dataset instead of reacting to every short online rollout.

Offline and online data distributions across three tasks.

Offline and online data coverage for selected laundry tasks.

Results

Average Success Rate

Average success rate comparison across laundry tasks.

DOU variants improve average success rate over the baseline across the evaluated laundry tasks.

Multi-Task Laundry Set

Multi-task laundry configurations.

Representative laundry configurations used for multi-task evaluation.

Unseen Tasks

Unseen laundry configurations.

Unseen laundry combinations for evaluating generalization.

Production Device

Production-ready washer product using offline reinforcement learning.

Production-ready target device with the offline RL control approach.

Videos

Supplemental motion example.

Naive rule-based baseline motion.

Proposed learned motion.

Poster

IEEE CAI 2024 poster.

Citation

@inproceedings{kang2024dataDrivenRLWasher,
  title     = {Data-Driven Reinforcement Learning for Optimal Motor Control in Washing Machines},
  author    = {Kang, Chanseok and Bae, Guntae and Kim, Daesung and Lee, Kyoungwoo and Son, Dohyeon and Lee, Chul and Lee, Jaeho and Lee, Jinwoo and Yun, Jae Woong},
  booktitle = {Proceedings - 2024 IEEE Conference on Artificial Intelligence (CAI)},
  pages     = {418--424},
  year      = {2024},
  doi       = {10.1109/CAI59869.2024.00083}
}