Waste sorting system using reinforcement learning

The waste classification system employs reinforcement learning to control robotic arms for adaptive waste sorting, addressing real-time challenges and enhancing accuracy and efficiency in waste classification.

WO2026127349A1 Publication Date: 2026-06-18AETECH

Patent Information

Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
AETECH
Filing Date
2025-10-24
Publication Date
2026-06-18

AI Technical Summary

Technical Problem

Existing waste sorting systems face challenges in accurately and efficiently classifying waste due to real-time fluctuations caused by factors such as waste type, size, and conveyor system status, requiring improved control of robotic arms to adapt to dynamic environments.

Method used

A waste classification system utilizing reinforcement learning to control the behavior of a robot arm, incorporating a sensor unit for data acquisition, a control unit for policy generation and action control, and a work unit with grippers to manage waste sorting tasks, with mechanisms for reward determination and policy updates based on state and cause information.

🎯Benefits of technology

Enhances classification accuracy and flexibility by adapting to unexpected situations, improving waste sorting efficiency and quality through continuous learning and feedback mechanisms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
Patent Text Reader

Abstract

According to the present invention, a waste sorting system using reinforcement learning comprises: a task unit for sorting a target object to be collected among waste moved by a conveyor device; a sensor unit for acquiring task data related to a sorting task of the task unit; and a control unit for analyzing the task data to generate state information related to the sorting task, performing reinforcement learning on the basis of the state information, and controlling an action of the task unit.
Need to check novelty before this filing date? Find Prior Art