Welcome!

Real-Time Embedded Systems

2024 – 2024

Energy-aware task scheduling for real-time embedded systems using Reinforcement Learning and DVFS

Project Overview

Designed an intelligent task scheduling framework for resource-constrained real-time embedded systems. The project leverages Reinforcement Learning (RL) and Dynamic Voltage and Frequency Scaling (DVFS) to minimize energy consumption while satisfying real-time execution constraints.

Key Features

  • Energy Optimization: Reduced power consumption through adaptive voltage and frequency scaling.
  • Reinforcement Learning: Developed an RL-based scheduler to learn energy-efficient scheduling policies.
  • Real-Time Scheduling: Ensured task deadlines and timing constraints were consistently satisfied.
  • Dynamic Resource Management: Optimized CPU performance based on workload characteristics and system state.
  • Performance Evaluation: Compared the proposed approach with conventional scheduling algorithms using simulation metrics.

Technologies Used

  • Python: Core implementation and experimentation.
  • Reinforcement Learning (RL): Intelligent scheduling policy optimization.
  • Dynamic Voltage & Frequency Scaling (DVFS): Energy-aware processor management.
  • Real-Time Systems: Task scheduling and deadline management.
  • NumPy & Matplotlib: Data analysis, simulation, and result visualization.

Software Architecture

  • Task Scheduler: Managed periodic and aperiodic real-time tasks.
  • RL Agent: Learned optimal scheduling decisions based on system states and rewards.
  • DVFS Controller: Dynamically adjusted processor voltage and frequency.
  • Simulation Environment: Evaluated scheduling performance under different workloads.
  • Analytics Module: Measured energy consumption, deadline miss rate, and overall system efficiency.

Impact and Applications

  • Energy-Efficient Computing: Extended battery life in embedded and IoT devices.
  • Real-Time Reliability: Maintained timing guarantees while minimizing power usage.
  • Embedded AI: Demonstrated the application of Reinforcement Learning in embedded system optimization.
  • Scalable Framework: Applicable to IoT devices, edge computing platforms, and battery-powered real-time systems.