• Context Recognition of Humans and Objects by Distributed Zero-Energy IoT Devices

  • By combining (i)zero-energy IoT devices, (ii)backscatter/RFID communication devices, and (iii)electronic circuits made by 3D printers, we will create zero-energy IoT devices for context recognition of humans and objects. In addition, we will build zero-energy IoT device networks combining these devices in mesh forms and create more advanced context recognition technology such as trajectory estimation and behavior recognition of humans and objects.

Passive IoT Device Networks

Various R&D activities are underway to realize a "super smart society" utilizing IoT, wireless communication, AI and big data. In order to realize a super smart society using IoT, it is important to spread battery-free and maintenance-free IoT devices (hereinafter referred to as “passive IoT devices”). Generally, IoT devices consume power for the three processes of sensing, process, and communication, but the power required for communication is extremely high (sensing is on the order of tens of μW, and wireless communication is on the order of several mW to hundreds of mW consumption), the key technology for the Internet connection of IoT devices is the spread of ultra-low power communication systems. In recent years, Wi-Fi-based backscatter communication technology (power consumption of about 10 μW) that can transmit and receive data at distances of several tens of meters at several Mbps and RFID communication technology that can transmit and receive data from a distance of several meters have been developed. In addition, a sensing element using only electric power obtained by energy harvesting and a low power consumption sensing technique for grasping human behavior have been devised.

However, existing sensors based on backscatter communication and RFID communication technologies are limited to the development of relatively simple

situation recognition technologies such as the presence or absence of a person at a target point. In this research, by utilizing the knowledge of the cross layer between the application layer and the physical layer in passive IoT device networks, and by building machine learning mechanisms using multiple passive IoT devices, we aim to create advanced situational awareness technology.

Situation Recognition of Humans and Things

In this research, by combining a battery-less passive IoT sensing device and an ultra-low power communication device such as backscatter communication, and utlizing an electronic circuit design technology using a 3D printer, we can develop custom-made passive IoT devices such as infrared sensors, acceleration sensors, cameras and thermometers that can be used to recognize situations of people and things. In addition, by buliding a passive IoT device network that combines many of these devices in a mesh form, we can achieve sophisticated situation recognition mechnisms such as estimating the movement trajectory of people and things and grasping activities, which is difficult to realize with a single passive IoT device. To create IoT device cooperation type situation recognition technology. Furthermore, using passive IoT devices that can be applied to the situation recognition of those people and things, we will create technology for (i) watching in

facilities for the elderly, (ii) grasping the activities of athletes, (iii) estimating moving trajectory of people and things, (iv) constructing sociograms for grasping children's human relationships, (v) grasping wind and ground fluctuations, (vi) air conditioning management of commercial facilities, and so on. We develop a design and development support environment for situation recognition systems using passive IoT device networks so that various situation recognition systems can be constructed.

Creating Energy-Harvesting Situation Recognition Technology

By developing various situation recognition systems using passive IoT device networks and evaluating and examining their effectiveness, we create situation recognition technologies that contribute to the realization of a "super smart society" promoted by the government. We believe that by realizing their design and development environment, it will contribute to the spread of various situation recognition systems.

[Selected Papers]

Y. Hara, R. Hasegawa, A. Uchiyama, T. Umedu, T. Higashino: "FlowScan: Estimating People Flows on Sidewalks Using Dashboard Cameras Based on Deep Learning", Journal of Information Processing, Vol.28, pp.55-64, 2020.

T. Higashino, A. Uchiyama, S. Saruwatari, H. Yamaguchi and T. Watanabe: “Context Recognition of Humans and Objects by Distributed Zero-Energy IoT Devices”, Proc. of 39th IEEE Int. Conf. on Distributed Computing Systems (ICDCS 2019), pp.1787-1796, 2019.

Y. Fukushima, D. Miura, T. Hamatani, H. Yamaguchi and T. Higashino: “MicroDeep: In-network Deep Learning by Micro-sensor Coordination for Pervasive Computing”, Proc. of 4th IEEE Int. Conf. on Smart Computing (SMARTCOMP 2018), pp.163-170, 2018.