Registration of new items in less than one-third the time by obviating re-training for existing items

Hitachi, Ltd. has developed a training technique for AI embedded in picking robots found in logistics warehouses. By applying this technique, the time required to register new items can be reduced to less than one-third that required by the conventional training technique.*1 With the conventional training technique, when a new item is introduced into a system, items already registered in the system need to be learned again together with the new item. Due to this need to train and validate all the items, greater time is required. The newly developed technique differs from the conventional technique in that it applies an incremental AI training technique which does not affect the recognition accuracy or speed of existing items and requires only for the new items to be learned. Hitachi intends to apply this technique as a component of Hitachi’s autonomous and collaborative robotics technology platform*2 to promote flexible and efficient automation of logistics operations.

This work is partially based on results obtained from a project, JPNP16007, commissioned by the New Energy and Industrial Technology Development Organization (NEDO).

画像: Comparison between conventional and proposed incremental training of AI

Comparison between conventional and proposed incremental training of AI

*1 Estimated based on the training time of a system that includes 10 items.
*2 'Development of Autonomous and Collaborative Robotics Technologies for Advanced Automation,’
Hitachi Review, vol. 68, no. 4, pp. 110-115, 2019. (PDF, 655kB)

For more information, use the enquiry form below to contact the Research & Development Group, Hitachi, Ltd. Please make sure to include the title of the article.

https://www8.hitachi.co.jp/inquiry/hitachi-ltd/hqrd/news/en/form.jsp

This article is a sponsored article by
''.