インダストリアルエンジニアリングとマネジメント

Automation and Robotics 2018: Motion spy- vibration energy harvesting sensor can track train passengers using machine learning techniques Marzieh Jalal Abadi- Data61, CSIRO, Australia

Abstract

Jalal Abadi

The present cell phones are outfitted with a scope of implanted sensors. These sensors can be utilized to induce relevant data, for example, area, action, wellbeing, and so forth and in this manner empower a scope of uses. Late exploration has shown that applications with access to information gathered from GPS, accelerometer and even gadget battery profile can precisely follow the area of clients as they move about in urban spaces. As of late, vibration vitality collecting (VEH) has developed as a reasonable choice for cell phones to address the deficiency of current battery innovation. VEH tackles power from human movements and encompassing sources and it could be utilized as a movement sensor. This is because of the way that distinctive surrounding vibrations and human movements produce a one of a kind example of vitality in the VEH circuit. In this paper, we uncover that VEH signal contains rich data and it is conceivable to decisively distinguish the excursion utilizing AI methods. A run of the mill train ride comprises of scenes of nonstop movement scattered with brief stoppages at train stations. Our key theory is that the train tracks between any two continuous stations make a novel vibration trademark that is reflected in the VEH information and we model it utilizing AI methods. At that point, we influence the consecutive idea of an excursion to address the incidental portion misclassifications and at last surmise the whole outing. To show our speculation, we gathered genuine movement information from 4 unmistakable train courses in the Sydney metropolitan region. Our informational collection incorporates movement information from 36 outings. To abuse a thresholding-based division calculation and concentrate the individual portions, we utilize diverse AI classifiers and group classifier accomplishes precision of 60.9% for distinguishing singular sections. At last, we utilize the successive properties of a train trip and accomplish an outing derivation precision of 97.2% for an excursion of 7 stations