Senior R&D Engineer — Hitachi Energy
2023 – present · Västerås, Sweden
- R&D of Hitachi's MACH control and protection DRTOS (C/C++, Yocto)
I'm an electrical engineer and computer scientist with a background in high-power systems. Over the past 20 years, I've worked on control and automation for converter stations and power substations. I have also spent time in academia, completing a Master's in Computer Engineering in Portugal with a focus on computer networks and advanced computing, followed by a PhD at ENAC/Onera/INSA in Toulouse researching distributed computing in dynamic networks.
I'm currently a Senior R&D Engineer at Hitachi, working as a System Architect on the DRTOS MACH platform used in control and protection systems across a wide range of industries and applications.
École doctorale Systèmes. Thesis: Modelling, Simulating and Emulating Distributed Applications in Swarms of Cyber-Physical Systems Deployed in Dynamic Networks.
École doctorale Systèmes. Thesis: Modelling, Simulating and Emulating Distributed Applications in Swarms of Cyber-Physical Systems Deployed in Dynamic Networks.
School of Engineering, Department of Informatics. Dissertation: Energy-Aware Gossip Protocol for Wireless Sensors Networks.
Centro de Informática. Field: Test Engineering for Embedded Systems.
Departamento de Engenharia Elétrica. Emphasis: Control and Automation.
The research field in distributed systems has witnessed a recent growing interest in wireless mobile distributed computing, i.e. distributed algorithms deployed in dynamic networks. Such systems introduce new challenges due to their time-varying topologies, which hinder the performance and safety guarantees of algorithms that are fundamental building blocks of more complex distributed algorithms. The main focus of this work is to establish different ways to study such systems via emulation and simulation and to propose some techniques to deploy applications in dynamic networks. We evaluated different data flow paradigms, such as many-to-many distributed storage, one-many computational offloading, and many-to-many decentralized swarm control and data load-balancing.
We adopted the quantitative approach as a cornerstone methodology, using simulation and emulation to collect data for analysis. There was a gap for emulation tools applied to mobile ad hoc computing — missing features included a full-scale emulated deployment environment where prototypes coexist with off-the-shelf applications, and richer mobility control. This work proposes MACE, a framework that enables the emulation of mobile distributed applications in a virtual environment so that scenarios and topologies composed by mobile wireless nodes can be easily modified. We also designed, implemented and validated a fluid model that simulates traffic flow in mobile ad hoc networks; the model can scale to hundreds of nodes and complete stress workloads in less than wall time.
Our work also encompassed proposals related to UTM, such as optimized edge-assisted offloading algorithms for swarms of UAVs (potentially used for UTM Distributed Detect and Avoid), and a distributed position tracking data layer for very-low-level airspace using State Machine Replication, achieving low end-to-end latencies even with many replicas.
In Wireless Sensor Networks (WSNs), typically composed of nodes with resource constraints, leveraging efficient processes is crucial to enhance network longevity and sustainability in ultra-dense, heterogeneous environments such as smart cities. Epidemic algorithms efficiently deliver packets but have poor energy efficiency due to redundancy. Directional algorithms (e.g. MCFA, Directed Diffusion) yield high efficiency but fail in mobile environments and have poor coverage.
This work proposes a new epidemic algorithm that uses the current energy state of the network to create a topology that is cyclically updated, fault tolerant, and able to handle the challenges of static or mobile heterogeneous networks. The protocol has a small computational footprint; required memory is proportional to a node's neighbour count rather than the network size, enabling high scalability. The protocol was tested using an ESP8266 as energy model reference in a simulated environment with ad-hoc wireless nodes, implemented at the application level over UDP, and proved highly energy efficient with delivery results comparable to a static directional algorithm.
In the near future, smart cities are expected to become more prevalent, with UAVs playing a key role in making cities more efficient and sustainable. Effective path planning is essential for safe and efficient integration of drones into urban airspace. UAVs may not have sufficient computing power to perform real-time contingency planning when encountering obstacles; this work proposes edge-assisted offloading scenarios where contingency planning is offloaded to nearby edge nodes. Various strategies were implemented and compared in a robot swarm simulator based on latency and cost metrics. Offloading plans generated using a genetic algorithm performed better in average latency or cost per offloading, with higher runtime overhead than other strategies.
Mobile CPS Swarms will likely change daily life in smart cities, space exploration, environmental monitoring, and transport. Industrial projects rely on centralized communication, leading to suboptimal performance and limited autonomy. To enhance distributed coordination within the swarm, reliable distributed computing through consensus protocols is a promising approach, yet only a handful of studies evaluate it in this context. We evaluate three key deployment aspects of mobile CPS swarms running consensus: cost of mobility, impact of node connectivity and partitions, and price of continuous routing updates. Results indicate transient partitions are common and that sparsity of dynamic networks plus continuous routing updates have a major impact on consensus performance.
Emerging edge applications introduced new computing time-variant topologies with mobile nodes connected via ad hoc networks, fundamentally different from cloud infrastructures. This work introduces an analytical model based on fluid quantities to study data flows of distributed systems in mobile ad hoc networks. Using an approach based on a network of queues with evenly distributed bandwidth over concurrent flows, this lightweight model enables fast, coarse-grained analysis of different distributed systems configurations and topologies. The model was implemented, validated and evaluated with stress workloads.
Integration of Unmanned Aircraft Systems into the airspace is a complex task. To ensure safe, secure UAS integration, operators and regulators should rely on a wide variety of dependable subsystems, including a tracking and position reporting system. This paper proposes a system architecture for tracking and position reporting of UAS in very-low-level airspace, leveraging low-latency communications and distributed computing for highly available, consistent tracking. Preliminary results suggest the approach is especially useful for traffic management in densely populated areas.
Designing and deploying mobile applications for Edge/Fog environments is challenging. Previous studies introduced simulation/emulation tools but fail to deliver an environment for emulating realistic deployments. We propose MACE, providing features that facilitate and accelerate development and testing of Edge/Fog applications, with mobility control that can be modelled or triggered by external applications such as a flight simulator. Two experimental evaluations are presented, plus the demonstration of mobility control via an external application.
In WSNs typically composed of resource-constrained nodes, leveraging efficient processes is crucial to enhance network lifetime in ultra-dense and heterogeneous environments. Balancing energy required to transport data efficiently across dynamic environments poses challenges to routing protocol design; reducing data redundancy while achieving acceptable delivery is a fundamental research topic. This work proposes a new energy-aware epidemic protocol that uses the current network energy state to create a dynamic distribution topology by self-adjusting each node's forwarding behaviour as eager or lazy according to local residual battery. Simulated evaluations demonstrate efficiency in energy consumption, delivery rate, and reduced computational burden compared with classical gossip protocols and a directional protocol.
Mother tongue: Portuguese (Brazil)
| Language | Listening | Reading | Spoken Interaction | Spoken Production | Writing |
|---|---|---|---|---|---|
| English | C2 | C2 | C2 | C1 | C1 |
| Swedish | B1 | B1 | B1 | B1 | B1 |
| Norwegian (Bokmål) | A2 | A2 | A2 | A2 | A2 |
| French | A2 | A2 | A2 | A2 | A2 |
Location: Västerås, Sweden