Carolynn Bernier

In this article, selected new directions in knowledge-based artificial intelligence (AI) and machine learning (ML) are presented: ontology development methodologies and tools, automated engineering of WordNets, innovations in semantic search, and automated machine learning (AutoML). Knowledge-based AI and ML complement each other ideally, as their strengths compensate for the weaknesses of the other discipline. This is demonstrated via selected corporate use cases: anomaly detection, efficient modeling of supply networks, circular economy, and semantic enrichment of technical information.

This article presents an instruction-set extension to the open-source RISC-V ISA (RV32IM) dedicated to ultra-low power (ULP) software-defined wireless IoT transceivers. The custom instructions are tailored to the needs of 8/16/32-bit integer complex arithmetic typically required by quadrature modulations. The proposed extension occupies only two major opcodes and most instructions are designed to come at a near-zero energy cost. Both an instruction accurate (IA) and a cycle accurate (CA) model of the new architecture are used to evaluate six IoT baseband processing test benches including FSK demodulation and LoRa preamble detection. Simulation results show cycle count improvements from 19 to 68 percent. Post synthesis simulations for a target 22nm FD-SOI technology show less than 1 percent power and 28 percent area overheads, respectively, relative to a baseline RV32IM design. Power simulations show a peak power consumption of 380 µW for Bluetooth LE demodulation and 225 µW for LoRa preamble detection (BW = 500 kHz, SF = 11).

Many Internet of Things applications are deployed over shared ISM (Industrial, Scientific, Medical) radiof requency spectrum bands. With the recent development of Low Power Wide Area (LPWA) wireless networks, the probability of interference and frame collisions has significantly increased. In this context, real-time interference monitoring is essential to provide precious information for network planning, base-station installation site selection, congested area detection, etc. This work presents a novel, low complexity interference detector for mobile LPWA nodes that is designed using a set of experimental data acquisitions. A first classifier is used to detect the presence of interference and shows a detection accuracy of 94%. If interference is detected, a second classifier is used to classify the interference’s relative strength into ten classes with a correct classification rate up to 97%. The detector also provides an estimation of the interference’s duration with an average relative error of 2% for medium to strong interference levels.

Current Low Power Wide Area Network (LPWAN) wireless transceivers are designed, or configured at deployment time, to function assuming a worse-case application scenario. Most of the time, they waste a significant amount of energy when operated under favourable channel conditions. Energy efficient and accurate channel state classification is imperative for selecting the optimum trade-off between transceiver performance and amount of saved energy, without impacting transmission quality. This work presents a novel, low complexity channel state indicator and a simple mono-feature classifier for channel state recognition. The classifier is trained using a set of experimental data acquisitions and has a 96% accuracy when tested with a new collected data set. Specially designed for LPWA applications, the classifier is capable of distinguishing undisturbed channels from those suffering from both mobility-induced fading and interference, while operating at low to medium SNR.

As of today, the machine-to-machine and Internet of Things (IoT) markets have significantly benefited from the deployment of diverse and efficient terrestrial low-power wide area networks (LPWAN) thanks to technologies such as SIGFOX UNB, LoRa, or NB-IoT. To further increase the coverage of such networks to remote areas, a satellite-based network with close interconnection to a terrestrial one is required. In this brief, we show the feasibility of a satellite link based on system and technological concepts similar to those used for LPWAN. We additionally prove the possibility of integrating the terminal transceiver in a single CMOS IC with power consumption and cost of the same order of magnitude as the terrestrial one. To this end, a complete L-band transceiver has been implemented. The transmitter modulates 400 symbols/s QPSK signals at as high as +12 dBm output power for 22.4 mA consumption under 3.3 V. The receiver consumes 4.7 mA while detecting the presence of low Earth orbit satellite pilots and demodulating the 20 ksymbols/s QPSK signal with -112 dBm sensitivity. The baseband processing handles up to 40 kHz Doppler shift and +/ - 440 Hz/s Doppler drift. The IC is processed in 65-nm CMOS technology from Taiwan Semiconductor Manufacturing Company Limited.

A complex wind-energy harvesting wireless sensor network WSN application for subway tunnels is pre-prototyped using the HarvWSNet co-simulation framework. Detailed models of every component, including energy harvester, tunnel propagation channel, train mobility, node power consumption, and physical layer PHY, medium access control MAC, routing and power aware protocols are described. Pre-prototyping simulations quickly identify the key parameters affecting the viability of the application e.g., harvested energy, train-induced signal attenuation. Thanks to its native capacity to handle detailed charge flow models, HarvWSNet can address scenarios with complex and intermittent energy supplies. This work demonstrates how co-simulation platforms such as HarvWSNet can aid in the development and future success of energy harvesting WSN applications.

International publishers of academic, scientific and professional journals since 1979.

This paper presents a reconfigurable receiver model whose purpose is to enable the study of reconfiguration strategies for future energy-aware and adaptive transceivers. This model is based on Figure of Merits of measured circuits. To account for real-life RF interference mechanisms, a link quality estimator is also provided.We show that adapting the receiver performance to the channel conditions can lead to considerable power saving. The models proposed can easily be implemented in a wireless network simulation in order to validate the value of a reconfigurable architecture in real-world deployment scenarios.