SensorLab at Jozef Stefan Institute

Time series anomaly detection is an important and yet unsolved challenge that is relevant for developing, maintaining and monitoring various aspects of smart infrastructures. One of our original contributions to the area consists in the definition of four types of anomalies relevant for monitoring the quality of the wireless links between various smart objects. A second original contribution consists of using time series to image transformations for advancing the state of the art in detection performance.

Wireless links are crucial to cost efficiently connecting various components in smart infrastructures. Recently, machine learning techniques proved to be suitable for more accurate estimation and classification. As original contribution to this area, we provide a comprehensive survey on link quality estimators developed from empirical data and then focus on the subset that use ML algorithms. We analyze ML-based Link Quality Estimation (LQE) models from two perspectives using performance data. Firstly, we focus on how they address quality requirements that are important from the perspective of the applications they serve. Secondly, we analyze how they approach the standard design steps commonly used in the ML community.

One important aspect of smart infrastructure research and development consists of automatic classification of time series data. Our recent work using shallow and deep machine learning methods for appliance classification in smart grids is documented in two recent papers and the code is available also as a CaaS service.

Location based services, already popular with end users, are now inevitably becoming part of new wireless infrastructures and emerging business processes. The increasingly popular Deep Learning (DL) artificial intelligence methods perform very well in wireless fingerprinting localization based on extensive indoor radio measurement data. Check out our work and datasets on improving ML based localization using traces from the LOG-a-TEC testbed and soon to come sustainable fingerprinting.

Reasearching MLOps pipelines and end-to-end machine learning development processes to design, build and deploy ML models.

Automation infrastructure for developing testing and deploying software for embedded/constrained devices is currently approaching market adoption. Our group pioneered continuous integration for embedded development by proposing the COINS framework and then implementing and evaluating it on LOG-a-TEC. We also touched on improving the efficiency of zero-touch device provisionining.

We developed the LOG-a-TEC experimental infrastructre that enables a wide range of outdoor and indoor experimentation such as with ultra narrow band and ultra wide band, packet based experimentation, clean slate protocol design, composable and modular protocol stacks, custom and advanced spectrum sensing and signal generating functions in sub-GHz spectrum.

We are commited to open science principles and we make available our early work as arxiv pre-prints, openly access(ible) versions of our accepted papers as well as datasets and code.

We offer various activities targeting education and knowledge transfer. Through our work with undergraduate, graduate and post-graduate students on individual cutting edge projects and their theses, we make sure they are better prepared for the market needs. We also organize summer schools, tutorials at conferences and locally in collaboration with the University of Ljubljana and the Jozef Stefan Internation Postgraduate School. For general practitioners and decision makers we also reccommend the edited book The Internet of Things: From Data to Insight.