Reducing Energy Consumption and Health Hazards of Electric Liquid Mosquito Repellents through TinyML.

Two problems arise when using commercially available electric liquid mosquito repellents. First, prallethrine, the main component of the liquid repellent, can have an adverse effect on the human body with extended exposure. Second, electricity is wasted when no mosquitoes are present. To solve these problems, a TinyML-oriented mosquito sound classification model is developed and integrated with a commercial electric liquid repellent device. Based on a convolutional neural network (CNN), the classification model can control the prallethrine vaporizer to turn on only when there are mosquitoes. As a consequence, the repellent user can avoid inhaling unnecessarily large amounts of the chemical, with the added benefit of dramatically reduced energy consumption by the repellent device.

An On-Device Deep Learning Approach to Battery Saving on Industrial Mobile Terminals.

The mobile terminals used in the logistics industry can be exposed to wildly varying environments, which may hinder effective operation. In particular, those used in cold storages can be subject to frosting in the scanner window when they are carried out of the warehouses to a room-temperature space outside. To prevent this, they usually employ a film heater on the scanner window. However, the temperature and humidity conditions of the surrounding environment and the temperature of the terminal itself that cause frosting vary widely. Due to the complicated frost-forming conditions, existing industrial mobile terminals choose to implement rather simple rules that operate the film heater well above the freezing point, which inevitably leads to inefficient energy use. This paper demonstrates that to avoid such waste, on-device artificial intelligence (AI) a.k.a. edge AI can be readily employed to industrial mobile terminals and can improve their energy efficiency. We propose an artificial-intelligence-based approach that utilizes deep learning technology to avoid the energy-wasting defrosting operations. By combining the traditional temperature-sensing logic with a convolutional neural network (CNN) classifier that visually checks for frost, we can more precisely control the defrosting operation. We embed the CNN classifier in the device and demonstrate that the approach significantly reduces the energy consumption. On our test terminal, the net ratio of the energy consumption by the existing system to that of the edge AI for the heating film is almost 14:1. Even with the common current-dissipation accounted for, our edge AI system would increase the operating hours by 86%, or by more than 6 h compared with the system without the edge AI.

Vehicle-to-Vehicle (V2V) Message Content Plausibility Check for Platoons through Low-Power Beaconing.

Although the IEEE Wireless Access in Vehicular Environment (WAVE) and 3GPP Cellular V2X deployments are imminent, their standards do not yet cover an important security aspect; the message content plausibility check. In safety-critical driving situations, vehicles cannot blindly trust the content of received safety messages, because an attacker may have forged false values in it in order to cause unsafe response from the receiving vehicles. In particular, the attacks mounted from remote, well-hidden positions around roads are considered the most apparent danger. So far, there have been three approaches to validating V2X message content: checking based on sensor fusion, behavior analysis, and communication constraints. This paper discusses the three existing approaches. In addition, it discusses a communication-based checking scheme that supplements the existing approaches. It uses low-power transmission of vehicle identifiers to identify remote attackers. We demonstrate its potential address in the case of an autonomous vehicle platooning application.

A Smartphone Magnetometer-Based Diagnostic Test for Automatic Contact Tracing in Infectious Disease Epidemics.

Smartphone magnetometer readings exhibit high linear correlation when two phones coexist within a short distance. Thus, the detected coexistence can serve as a proxy for close human contact events, and one can conceive using it as a possible automatic tool to modernize the contact tracing in infectious disease epidemics. This paper investigates how good a diagnostic test it would be, by evaluating the discriminative and predictive power of the smartphone magnetometer-based contact detection in multiple measures. Based on the sensitivity, specificity, likelihood ratios, and diagnostic odds ratios, we find that the decision made by the smartphone magnetometer-based test can be accurate in telling contacts from no contacts. Furthermore, through the evaluation process, we determine the appropriate range of compared trace segment sizes and the correlation cutoff values that we should use in such diagnostic tests.

Reducing Message Collisions in Sensing-Based Semi-Persistent Scheduling (SPS) by Using Reselection Lookaheads in Cellular V2X.

In the C-V2X sidelink Mode 4 communication, the sensing-based semi-persistent scheduling (SPS) implements a message collision avoidance algorithm to cope with the undesirable effects of wireless channel congestion. Still, the current standard mechanism produces a high number of packet collisions, which may hinder the high-reliability communications required in future C-V2X applications such as autonomous driving. In this paper, we show that by drastically reducing the uncertainties in the choice of the resource to use for SPS, we can significantly reduce the message collisions in the C-V2X sidelink Mode 4. Specifically, we propose the use of the "lookahead", which contains the next starting resource location in the time-frequency plane. By exchanging the lookahead information piggybacked on the periodic safety message, vehicular user equipment (UEs) can eliminate most message collisions arising from the ignorance of other UEs' internal decisions. Although the proposed scheme would require the inclusion of the lookahead in the control part of the packet, the benefit may outweigh the bandwidth cost, considering the stringent reliability requirement in future C-V2X applications.

Sensorless PV Power Forecasting in Grid-Connected Buildings through Deep Learning.

Existing works in photovoltaic (PV) power generation focus on accurately predicting the PV power output on a forecast horizon. As the solar power generation is heavily influenced by meteorological conditions such as solar radiation, the weather forecast is a critical input in the prediction performance. However, the weather forecast is traditionally considered to have coarse granularity, so many are compelled to use on-site meteorological sensors to complement it. However, the approach involving on-site sensors has several issues. First, it incurs the cost in the installation, operation, and management of the sensors. Second, the physical model of the sensor dynamics itself can be a source of forecast errors. Third, it requires an accumulation of sensory data that represent all seasonal variations, which takes time to collect. In this paper, we take an alternative approach to use a relatively large deep neural network (DNN) instead of the on-site sensors to cope with the coarse-grained weather forecast. With historical PV output power data from our grid-connected building with a rooftop PV power generation facility and the publicly available weather forecast history data, we demonstrate that we can train a six-layer feedforward DNN for the day-ahead forecast. It achieves the average mean absolute error (MAE) of 2.9%, comparable to that of the conventional model, but without involing the on-site sensors.

Empirical Determination of Efficient Sensing Frequencies for Magnetometer-Based Continuous Human Contact Monitoring.

The high linear correlation between the smartphone magnetometer readings in close proximity can be exploited for physical human contact detection, which could be useful for such applications as infectious disease contact tracing or social behavior monitoring. Alternative approaches using other capabilities in smartphones have aspects that do not fit well with the human contact detection. Using Wi-Fi or cellular fingerprints have larger localization errors than close human contact distances. Bluetooth beacons could reveal the identity of the transmitter, threatening the privacy of the user. Also, using sensors such as GPS does not work for indoor contacts. However, the magnetometer correlation check works best in human contact distances that matter in infectious disease transmissions or social interactions. The omni-present geomagnetism makes it work both indoors and outdoors, and the measured magnetometer values do not easily reveal the identity and the location of the smartphone. One issue with the magnetometer-based contact detection, however, is the energy consumption. Since the contacts can take place anytime, the magnetometer sensing and recording should be running continuously. Therefore, how we address the energy requirement for the extended and continuous operation can decide the viability of the whole idea. However, then, we note that almost all existing magnetometer-based applications such as indoor location and navigation have used high sensing frequencies, ranging from 10 Hz to 200 Hz. At these frequencies, we measure that the time to complete battery drain in a typical smartphone is shortened by three to twelve hours. The heavy toll raises the question as to whether the magnetometer-based contact detection can avoid such high sensing rates while not losing the contact detection accuracy. In order to answer the question, we conduct a measurement-based study using independently produced magnetometer traces from three different countries. Specifically, we gradually remove high frequency components in the traces, while observing the correlation changes. As a result, we find that the human coexistence detection indeed tends to be no less, if not more, effective at the sampling frequency of 1 Hz or even less. This is because unlike the other applications that require centimeter-level precision, the human contacts detected anywhere within a couple of meters are valid for our purpose. With the typical smartphone battery capacity and at the 1 Hz sensing, the battery consumption is well below an hour, which is smaller by more than two hours compared with 10 Hz sampling and by almost eleven hours compared with 200 Hz sampling. With other tasks running simultaneously on smartphones, the energy saving aspect will only become more critical. Therefore, we conclude that sensing the ambient magnetic field at 1 Hz is sufficient for the human contact monitoring purpose. We expect that this finding will have a significant practicability implication in the smartphone magnetometer-based contact monitoring applications in general.