- Operation and Control
Two-stage control learning technology for image-based control
Reduces work in the application of controls and systems through end-to-end learning that does not require design of image recognition processing.
- Learns policy for controlling robots from input images.
- Reduces labor by learning control policy from accumulated operation data by experts or simulation environments.
- Achieves high accuracy control by learning a control policy and a policy for correcting the control simultaneously.
Applications
- Automation systems in logistics, manufacturing, maintenance, and inspection facilities that use robots
- Automated control in moving bodies (e.g., AGVs and drones)
- Control assistance for defense, medical devices, etc.
Benchmarks, strengths, and track record
- The control policy is not susceptible to the surrounding environments, and can be easily learned from simulation environments.
- When simulation environments are not available, control policy can be learned from accumulated operations data by experts.
Inquiries
Inquiries to Toshiba Corporate Laboratory (Komukai region)
Please include the title “Toshiba AI Technology Catalog: Two-stage control learning technology for image-based control” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
References:
- Q. Dong, et al., "An Offline Learning of Behavior Correction Policy for Vision-Based Robotic Manipulation", In 41st IEEE Conference on Robotics and Automation (ICRA 2024).
- Toshiba developed the top precise control of complex robot operations using ‘offline reinforcement learning’ with a small amount of data -training AI safely and efficiently for promoting automation in manufacturing sites, contributing to the alleviation of labor shortages-. R&D Center; Toshiba (global.toshiba)
