成果報告書詳細
管理番号20160000000576
タイトル*平成27年度中間年報 次世代ロボット中核技術開発 次世代人工知能技術分野 人間と相互理解できる次世代人工知能技術の研究開発  
公開日2016/7/27
報告書年度2015 - 2015
委託先名国立研究開発法人産業技術総合研究所
プロジェクト番号P15009
部署名ロボット・AI部
和文要約
英文要約Title: Next Generation Robot Core Technology Research and Development Project/
Next Generation Artificial Intelligence Technology Area (Feasibility Study) (FY2015-FY2016) FY2015 Annual Report

Aiming to realize the world where human intelligence and artificial intelligence (AI) can cooperate for solving challenging social issues, we are investigating the next generation artificial intelligence technologies which enhance mutual understanding between humans and AI systems. Major progresses in FY 2015 are as follows;
Next generation brain-inspired AI: theoretical confirmation and parallel distributed implementation of BESOM, designing visual and language area using BESOM, proposing methods of estimating spatial STA and temporal STA of visual area neurons, developing a model of basal-ganglia and cerebellum, and proposing an object tracking method with active learning.
Data-knowledge integration AI: selecting caption generation and Q&A regarding image/video and economic data as target problems, and proposing a method for transforming knowledge in distributed representation to symbolic representation.
Advanced machine learning and probabilistic modeling: investigating methods of data compression for scalable learning, proposing data-augmentation methods for learning with highly complex models, implementing a probabilistic programming language, and investigating methods for combining deep learning with reinforcement learning.
Next generation AI framework: prototyping an AI framework and common data models, procuring a cluster for AI research, providing resources of cloud computing, and collecting data of human-robot interactions in simulated environments.
Advanced AI core modules: constructing “living lab” for collecting data of everyday activities, collecting 3D shape data of 50 commodities, proposing methods of measuring/planning assembly motions, preparing experiments of motion imitation and flexible object handling, and investigating natural language processing modules.
Modeling human behaviors: investigating target fields and preparing systems for collecting data (such as next generation bending machines), measuring behaviors, and acquiring knowledge.
Image analysis: constructing a data base of satellite images and evaluating the performance of change detection and object detection modules.
Incident report analysis and accident prevention: designing a format for describing processes of accidents and analyzing incident data.
Human-machine interaction: constructing a sensor system for identifying and tracking children in kindergartens and conducting interaction experiments.
Industrial robot control: constructing an industrial robot system and a picking simulator for picking learning.
Autonomous driving car: pre-processing incident data for risk prediction learning, constructing advanced driving support system ontology, and starting to design hardware systems for real-time processing.
ダウンロード成果報告書データベース(ユーザ登録必須)から、ダウンロードしてください。

▲トップに戻る