Building Adaptive, Dependable, and Secure Systems with a Cross-Layer Weakly-Hard Paradigm
Abstract
Research Problem: Many embedded real-time systems, such as those in robotic vehicles and industry automation, are bounded by resource constraints on computation, communication, energy, etc. During their design and operation, there are often strong needs to adapt them under changing environment and retrofit them with new features (e.g., to strengthen security protection and faulttolerance). However, conducting such online adaptation and offline changes is often challenging and sometimes infeasible due to stringent resource and timing constraints. A fundamental limitation comes from the current paradigms of enforcing system timing behavior ~ rigid and overly pessimistic hard deadlines make it difficult to carry out changes, whereas simply allowing soft deadlines cannot provide the assurance on performance and dependability that mission-critical systems often demand.This project will investigate a weakly-hard paradigm for defining timing constraints and guiding system adaptation and changes. This is motivated by the fact that many system functions can tolerate certain degrees of timing violations and still provide guarantees on functional and extrafunctionalproperties such as safety, stability and performance, as long as these violations are bounded and controlled. Leveraging weakly-hard constraints can significantly expand the feasible system configuration space and thus enable adaptation and changes that could not have been achieved with hard deadlines, while providing deterministic and quantifiable guarantees on system properties and assurance on the overall system dependability.Objective and Approach: The objective of this project is to develop a cross-layer weakly-hard framework that enables dependable, nimble and robust online adaptation and offline changes. The framework includes three closely-knitted research thrusts:Thrust 1: Functional Verification with Weakly-Hard Constraints. This thrust will develop formal methods to analyze the impact of weakly-hard constraints on critical functional and extrafunctional properties, and verify the feasible set of weakly-hard constraints for system functions in sensing, computation, control or communication. Intuitively, the goal is to determine how many deadline misses (and in what manner) each function can tolerate while still meeting its functionaland extra-functional requirements.Thrust 2: Software Synthesis and Co-design/Co-adaptation in Weakly-Hard Systems. This thrust will develop efficient algorithms to optimize the software architecture in weakly-hard systems. The goal is to maximize the timing and resource ~slack~ with respect to the given weakly-hardconstraints (verified by Thrust 1), for best meeting system adaptation and change needs. This thrust also investigates how to formulate and set weakly-hard constraints in the first place to trade off different system objectives.Thrust 3: System Support for Weakly-Hard Dependability. Software tasks with weakly-hard constraints (verified and set by Thrusts 1 and 2) need system~s support to maintain their functional and temporal correctness during execution. This thrust will develop system primitives and abstractions for executing weakly-hard tasks in an efficient and dependable manner, including new job-class schedulers for better resource utilization and predictability and runtime mechanisms for safe and robust operations of deadline-missed jobs.Outcome and Impact: The proposed framework will lay the foundation for enabling dependable, nimble and robust online adaptation and offline changes to mission-critical systems that operate in dynamic environment contexts. It will bring significant benefits to such systems, e.g., by strengthening their security and fault tolerance in adversarial environments, by improving controlperformance in difficult-to-navigate surroundings, and by accommodating software updates for adding new features or fixing bugs.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 20, 2019
- Source ID
- N000141912496
Entities
People
- Qi Zhu
Organizations
- Northwestern University
- Office of Naval Research
- United States Navy