Email
DNINFOA - SIA
Libraries
Calls
UNAL identity
Research line
Dependable embedded systems
We develop techniques and tools for designing fault-tolerant hardware/software systems.
How can an embedded system keep operating correctly when its hardware suffers transient faults? At PLaS we research methods and tools to improve the dependability of microprocessor-based systems, with special attention to radiation-induced errors. We seek to balance fault protection against the memory, hardware area, energy and execution-time constraints typical of these systems.
We combine software- and hardware-implemented protection techniques, criticality analysis and approximate computing. We develop and evaluate fault-injection tools and predictive models that make it possible to compare design alternatives and guide the choice of fault-tolerance mechanisms from the early stages of development.
Topics
Fault tolerance23 works
We design and evaluate mechanisms to detect and mitigate transient errors that can corrupt a program's data or control flow. We work with redundancy, compiler-driven code transformations and hardware/software co-design strategies, on both commercial microprocessors and FPGA-based processors. We aim to apply protection selectively, according to each application's needs, and to measure its effectiveness and cost.
Approximate computing for dependable systems7 works
We investigate how to exploit the tolerance of some applications to small variations in their results in order to reduce the cost of protection mechanisms. We combine approximate versions of computations with redundancy and selective protection techniques, and evaluate the trade-off between output quality, dependability and resource usage. We study when these strategies are useful and which approximation limits still meet the application's requirements.
Dependability assessment8 works
We develop tools and methods to generate and compare protection alternatives through fault-injection campaigns, simulation and experiments on real processors. We analyze the fidelity of these assessments and how well they anticipate system behavior. We also research machine learning models that estimate dependability early and support design space exploration.
Timeline
2026
2024
2023
2022
2021
2019
- MiFIT: A Fault Injection Tool to Validate the Reliability of Microprocessors
- Multi-Threaded Mitigation of Radiation-Induced Soft Errors in Bare-Metal Embedded Systems
- Reducing Overheads in Software-based Fault Tolerant Systems using Approximate Computing
- Softerror mitigation for multi-core processors based on thread replication
- Using Approximate Computing and Selective Hardening for the Reduction of Overheads in the Design of Radiation-Induced Fault-Tolerant Systems
2018
2017
2016
- A Hardware-Software Approach for On-Line Soft Error Mitigation in Interrupt-Driven Applications
- Dependability evaluation of COTS microprocessors via on-chip debugging facilities
- Reducing Implicit Overheads of Soft Error Mitigation Techniques Using Selective Hardening
- Soft Error Mitigation in Soft-Core Processors
2015
2014
2013
2012
- A Co-Design Approach for SET Mitigation in Embedded Systems
- Actas de las XII Jornadas de Computación Reconfigurable y Aplicaciones 2012
- Compiler-Directed Soft Error Mitigation for Embedded Systems
2011
- A co-design approach for SET mitigation in embedded systems
- A Novel Co-Design Approach for Soft Errors Mitigation in Embedded Systems
- Co-diseño de sistemas hardware/software tolerantes a fallos inducidos por radiación
- On the definition of real conditions for a fault injection experiment on embedded systems
- Soft core based embedded systems in critical aerospace applications
2010
- A compiler-based infrastructure for fault-tolerant co-design
- Application-driven co-design of fault-tolerant industrial systems
- Application-Driven Co-design of Fault-Tolerant Industrial Systems
- Hardening Development Environment for Embedded Systems
- Rapid Prototyping of Radiation-Tolerant Embedded Systems on FPGA
2008
- Platform for intelligent management of Industrial Machinery based on service-oriented architecture