Search

Molecular communication via diffusion (MCvD) utilizes the release, diffusion and reception of molecules to transmit information. The receiver localization in vessel-like and flow-induced MCvD system has promising prospects in the field of drug delivery. The existing work mainly focuses on the receiver location using deep neural networks (DNN) model in two-dimensional and unbounded scenarios without flow. In this letter, considering a three-dimensional vessel-like MCvD system with Poiseuille flow, we employ the transformer-based model with three encoders to predict the location of the receiver. The data generation is built based on the time series of the numbers of information molecules received by the receiver from multiple transmitters. Numerical results show that the transformer-based model exhibits better accuracy of prediction for receiver location in contrast to DNN-based model.

The mobile molecular communication (MMC) system has promising prospects in the field of biomedical drug delivery. The signal detection plays significant roles in improving the performance of the MMC system with time-varying channel. However, when the channel state information is unknown, traditional detection strategies are not effective and the detectors using deep learning can exhibit better detection ability. This letter proposes a signal sequence detector for the MMC system based on the Informer model which is one of the deep learning-based detectors. The computation of autocorrelation coefficient of the input sequence is utilized to find the optimal sequence length. Based on the fixed length of the input sequence, we simplify the structure of the Informer model to reduce the inference time. Numerical results demonstrate that the detection ability of the Informer-based detector outperforms the Transformer-based detector for signal sequence detection in the MMC system.

Commodity RFID tags backscatter stored electronic product code (EPC) to the reader, but do not have sensing capability. Existing works have made much effort on designing RFID-based sensing platform. But most of them either need intricate hardware design or rely on modification of the tag, which increases the cost or constrains the sensing capability. In this paper, we design a self-jamming identification and sensing platform (SJISP) consisting of SJISP nodes and a commodity RFID reader. A subtle design of the SJISP node is the adoption of a jammer radio module with the same frequency as the reader, controlled by the micro control unit (MCU) to decide whether to interfere with the query process of the RFID reader. The RFID tag is not readable if the jammer is turned on to generate interference signals. Otherwise, it is readable when the jammer is turned off. The sensing data is thus modulated by switching the jammer on and off for transmitting bit 0 and bit 1, respectively. The reader demodulates the data through the compatible EPC UHF Gen2 air interface protocol. To further save the energy of the SJISP node, we propose a prefix codebook based data delivery scheme, which leverages the difference of energy consumption (DEC) between transmitting bit 0 and bit 1. Our proposed scheme can save more than 50$\%$% of the energy than common communication without codebook. Experimental results based on our prototyped system show that the designed SJISP can achieve an average packet reception rate (PRR) of over 99$\%$% and is quite robust to environmental disturbance. Our designed platform provides a low-cost and compatible solution to extend the sensing capability of RFID system. A demo application with a temperature sensor and a light sensor embedded in two SJISP nodes respectively are developed to demonstrate how SJISP applies in real world scenario.

The risk of death and injury from traffic crashes has been universally recognized as one of the most serious threats to sustainable development. Among all the factors in traffic crashes, aggressive driving in which the driver usually makes excessive lane changes to overtake other vehicles is prevalent. As such, monitoring lane changes and providing real-time warnings is beneficial for improving transportation sustainability. This article presents BackWatch, a novel vehicle-mounted sensing system that uses a back view cabin camera monitoring the steering wheel rotations to track lane-change events. BackWatch consists of an encoder network to extract essential visual features of steering wheel rotations, and an inference network incorporating the visual and GPS speed features to recognize the resulting lane changes. Our system does not rely on precise coordinate alignment between the monitoring device and the vehicle, nor the wearables worn by the driver, and is robust against different drivers, vehicles, driving speeds, and environmental settings. We evaluate the system based on 16 hours of real-world on-road driving data collected from three pairs of cars and drivers under different traffic and environmental conditions. The results show that BackWatch achieves 0.952 of precision and 0.981 of recall on the detection of lane changes.