Simeio: Anomaly Detection for Building Automation System

Buildings are an important energy consumer and are equipped with hundreds of sensors and control systems. The analysis of such massive data can reveal insights for building owners to optimize the building infrastructure. Currently, usage of such data is limited to traditional control systems, energy commissioning, and maintenance on a regular basis.

Analytics on 5G – Topology through PM correlation

As the 5G Network will be capable of being reconfigured and optimized on-the-fly, they will also be more automated, requiring less manual effort to provision resources and make the most efficient use of bandwidth. The Ciena Analytics as a Service is a suite of tools to assist network operator in pro-active discovery on their network operations. This project will look at integrating new advanced intrepretable Artificial Intelligence and Machine Learning techniques to tackle different challenging networking tasks (e.g. traffic prediction, anomaly detection, topology discovery, etc.).

“Minimally invasive mechanical circulatory support scientific data acquisition”

Heart failure is a prevalent disease affecting 250,000 people in North America alone. This disease can be treated by the transplant of a donor organ, but insufficient donor organs have led to the development of mechanical circulatory support which now provide a reliable alternate treatment option for patients. Unfortunately, many patients that could be helped by a mechanical circulatory support are deemed ineligible due to the invasive, open- heart surgery that is required to install such devices. Puzzle Medical Inc.

Energy reduction in HVAC systems in a commercial building environment using data-driven approaches

The main goal of this project is to develop data-driven approaches to reduce energy consumption and cost when operating commercial building’s cooling systems. Indeed, according to recent studies the building sector is one of the largest energy-consuming entities (almost 40% of global energy consumption) and this consumption is predicted to increase by 50% by 2050. Thus, there is an urgent need to provide solutions to reduce energy consumption taking into account the importance to improve environmental sustainability and the increase of electricity prices.

Complementary and competitive interactions between wild and managed bees

A diversity of native bee species inhabit agricultural and urban landscapes and can be more effective pollinators than the widely employed European honey bee. However, honey and wild bee communities often overlap, which means these bees compete for the same floral resources. Studies of competition between wild and managed pollinators are limited due to methodological constraints. This restricts our ability to predict how pollination and bee diversity will be affected by changes in pollinator community composition.

Private SQL interface for encrypted data

Querying databases without a layer of privacy protection might lead to serious privacy issues. Such issues include access patterns and communication volume patterns. By combining the state-of-the-art privacy standard (differential privacy) and encryption in provides resilience to a host of attacks on remote databases, including data reconstruction attacks. However, there is still research work needed in building a private access system on top of an encrypted database.

D2K+: Deep Learning of System Crash and Failure Reports for DevOps

The objective of this project is to develop techniques and tools that leverage artificial intelligence to automate the process of handling system crashes at Ericsson, one of the largest telecom and software companies in the world, and where the handling of crash reports (CRs) and continuous monitoring of key infrastructures tend to be particularly complex due to the large client base the company serves. In this project, we will explore the use of deep learning algorithms to classify CRs based on a variety of features including crash traces, CR descriptions, and a combination of both.

Design and development of techniques to characterize optical, mechanical and chemical properties of metallic and semiconductor thin films with applications in MEMS structures and their packaging

Micro-Electro-Mechanical Systems (MEMS) are complex systems with sizes in the range of few microns (human hair has thickness of 150-200 microns) which have both mechanical and electronic components. MEMS technology has entered in many industries such as optical technology, point of care diagnostics, telecommunications, automotive, and military. Today, there are hundreds of MEMS devices, e.g. microscale gyroscopes and accelerometers, used in cars to control different components, including wheels, brakes, steering, and air bags.

Machine Learning and Data Mining Approaches for Smart Buildings

The goal of this project is to develop machine learning and data mining algorithms relying on non-intrusive common sensor data to estimate and predict smart buildings’ occupancy and activities. Efficient feedbacks are automatically supplied to the end user to involve occupants and increase their awareness about energy systems. This consists of generating reports helping the occupant to understand his/her energy management system and thus to be involved in the decision-making process.

Self-Adaptive Penetration Tests with Deep-Reinforced Intelligent Agents

Penetration testing is a key security tactic, where defenders thinks like an attacker to predict the latter’s actions and develop effective defense. However, for large-scale cyber-physical infrastructures like the smart grid, traditional penetration tests on individual devices or networks are insufficient to exhaust all potential exploits or to reveal infrastructure-level vulnerabilities invisible to the local system.