Semi-Supervised Deep Learning Techniques applied to Traffic Scene Understanding
Semantic Segmentation methods play a key role in today’s Autonomous Driving research, since they provide a global understanding of the traffic scene for upper-level tasks like navigation. However, main research efforts are being put on enlarging deep architectures to achieve marginal accuracy boosts in existing datasets, forgetting that these algorithms must be deployed in a real vehicle with images that weren’t seen during training. On the other hand, achieving robustness in any domain is not an easy task, since deep networks are prone to overfitting even with thousands of training images.
This research line proposes to analyze different techniques to be applied to existing deep networks in order to improve their robustness when deployed in any domain. Our proposal is to analyze semi-supervised methods in this context with the aim of implementing real applications for autonomous vehicles.
Remote gas sensing in Energy-production infrastructures
The development of systems for remote sensing is relevant to several applications such as leak detection in industry. Ultrafast fiber lasers represent a reliable source for this application, as they show a stable and with wide-spectrum emission. This Project aims to apply nitride-based ultrafast lasers developed by the Photonics Engineering Group of the University of Alcalá to remote gas detection in energy-production industry. These lasers attains high peak power (in the range of tens KW) linked to ultrashort pulse width (below 200 fs), emitting a powerfull pulsed beam spectraly centred at 1.56 µm, where gases like NH3, CO, CO2 and H2S show important absorption lines.
Integration of Wireless Sensor Networks and Event-based Sensing for Smart Cities applications
Geintra research group (UAH) has been working on wireles sensor network (WSN) since last years, and specifically on event-based sensing and estimation. That means sensorial nodes acitvate the transmission process, updating information, only when required. This way the communication channel is relaxed and an efficient power consumption is achieved without significant degradation of the supervision, monitoring or control performance. The challenges of this proposal are focused on the integration and contributions of event-based sampling to smart sensing appplied to some of the branches of Smart Cities: energy and environment, buildings and infraestructures, mobility and intermodality, governance and social services.
Network control and sensig for cooperation between mobile robots
The proposal address to the contribution of networked control systems, wireless sensor networks and industry 4.0 to the cooperation between autonomous guided vehicles (AGVs). The research lines should be related to: processing techniques that optimise power autonomy, event-based network control techniques, localization and mapping algorithms, visual-servoing and intelligent guiding systems.
Non-intrusive load monitoring and deep learning applied to Ambient Assisted Living
NILM and smart meters imply a new approach to monitor elderlies by using one single sensor, enabling scalability and non-intrusiveness. The energy disaggregation and appliance identification are actually the foundations, where a deep learning approach can propose novel methods to identity, determine and analyse behaviours, patterns and routines in the daily activities of people living in the households under analysis. These methods can provide relatives and carers with a powerful tool to evaluate and/or infer a person’s situation with a null intrusiveness over time, not only in the short but also in the long term.
Low power System-on-chip – Novel Design Techniques
This research topic addresses the configuration and study of low power system-on-chip (SoC) used in embedded systems where the autonomy is a key factor of success for the validity of any commercial project. We have experience in the research of FPGAs and system on chip modules based on Xilinx Zynq. The project might provide a common framework to compare the power consumption of embedded system and offering end-users with debugging capabilities to know where to put the efforts while designing for a low-power programable system on a SoC module.
Wearable Sensors for advanced energy management
This research topic addresses the deployment of wearable sensors that together with fixed nodes might be used in advanced energy management. We have experience in the application of hybrid systems for in indoor localization of people in smart buildings. The fixed IoT nodes provide an accurate positioning information and can be used as routers for uploading information to the cloud. Simulation of edge-devices plus the indoor localization techniques in IoT motes for a large deployment inside a smart building might be considered valuable.
Efficient Shared of Green Energy Consumption on Smart Buildings
Efficient Shared of Green Energy Consumption on Smart Buildings
This research topic addresses the deployment of IoT nodes on a smart building equipped with a shared renewable photovoltaic generation system. Collaborative economy and shared resources are in the background of this research line. Each flat owner will have IoT fixed nodes at the main consumer devices, requesting its service when best available. A credit system for each community should be envisaged so both the shared of green energy and efficiency are considered when connecting the devices. The IoT deployment shall be capable to control the electricity consumed by the equipments choosing the best time to be switched on lowering the electricity demand while maximizing the consumption of all the available generated green electricity. The goal is to match the generation of the photovoltaic system (green energy) with the demand so the community takes advantage of the whole green electricity generation. The IoT network would keep track of the delayed requests done by users, switching on the equipments based on a smart schedule, while reporting to everyone their consumption so the people behaviour could be changed to better use the energy.
Fast Bidirectional Electrical Vehicle Chargers
Increase in electric vehicle mobility has encouraged the growth of vehicle to grid technology. Vehicle to grid technology allows bidirectional power flow between the battery of electric vehicle and the power grid. In this project, a new control strategy for bidirectional battery charger is proposed. The proposed control strategy can charge and discharge an electric vehicle battery in fast mode. Also, topologies of the converter including renewable energies will be included to obtain very fast chargers
Grid supporting using virtual power plants based on aggregated local networks
A virtual power plant (VPP) is a cloud-based distributed power plant that aggregates the capacities of heterogeneous Distributed Energy Resources (DERs) for the purposes of enhancing ower generation, as well as trading or selling power on the open market. The main aim of this project is to use the aggregated capabilities of the a lot of systems to participate in the grid supporting from the point of view of voltage and frequency.