成果報告書詳細
管理番号20160000000329
タイトル*平成27年度中間年報 エネルギー・環境新技術先導プログラム 究極の省エネを実現する「完全自動化」自動車に不可欠な革新認識システムの研究開発
公開日2016/5/17
報告書年度2015 - 2015
委託先名一般財団法人マイクロマシンセンター 株式会社デンソー 国立大学法人東京大学
プロジェクト番号P14004
部署名イノベーション推進部
和文要約
英文要約Title: Advanced Research Program for Energy and Environmental Technologies / Innovative Recognition Systems for autonomous driving (FY2014-FY2016) FY2015 Annual Report

Final goal is to realize ultimately energy-effective transportation system which cuts CO2 emission by 25% via developing innovative technologies for autonomous driving, which is impossible to be developed with the current technology. These innovative technologies are divided into three issues: (1) development of a molecular inertial gyroscope such that a car always grasps its exact location, (2) development of a spectroscopic imager for accurate and regular recognition of the car’s surroundings, and (3) development of an algorithm of visual recognition with high accuracy. This project evaluated feasibility of these three unprecedented technologies by taking actual driving environments into consideration, and obtained these results described below.
(1)Molecular inertial gyroscope
We confirmed a feasibility of driftless gyroscope operation through an investigation on possible causes of drift. In addition, we performed fabrication step tests of sub-micron-thick cantilever to check applicability to MEMS 8-inch production line. We also clarified spec requirements of the gyroscope as an on-vehicle sensor by taking actual environmental conditions into account, and constructed an experimental instrument setup to evaluate the gyroscope performance. We also improved estimation of the amount of the energy consumption cut by this technology.
(2)Spectroscopic imager
Aiming at realizing infrared light detection using silicon as a material, we clarified the design principle of the nano antenna for infrared light absorption. We could electrically detect the near-infrared light by a barrier structure. Strategies for realizing a target spec concerning noise equivalent temperature difference (NETD) was also clarified. These outcomes were presented in both domestic and international conferences. As an applicability check to the semiconductor production line, we performed fabrication step tests of the device whose designs were based on results described above. The light detection sensitivity was confirmed to increase using the device. We also clarified spec requirements of the imager as an on-vehicle sensor by taking actual environment into consideration, and constructed a setup to evaluate the sensor performance such as NETD.
(3)Algorithm of visual recognition with high accuracy
We constructed a practical pseudo-coaxial camera, and prepared image data composed of sets of visible and far-infrared human figures. The data was used for deep learning, and it was confirmed that the performance of human discrimination was improved using this multi-wavelength data. In addition, we clarified actual environment where our imager works effective as an on-vehicle sensor by considering target objects to be recognized.
As results of this project, we made five public presentations. Because all of the interim goals set at the start of this project were achieved, this project passed a stage gate judgement.
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