https://github.com/amirreza81/rl-dvfs-scheduling
Intelligent Power Management & Task Scheduling for Real-Time Embedded Systems | Reinforcement Learning meets DVFS for Maximum Efficiency
https://github.com/amirreza81/rl-dvfs-scheduling
dvfs embedded-systems energy-optimization iot multi-core power-management qos-optimization real-time-scheduling reinforcement-learning rl rl-based-scheduling task-scheduling tdp thermal-design-power thermal-management
Last synced: 2 months ago
JSON representation
Intelligent Power Management & Task Scheduling for Real-Time Embedded Systems | Reinforcement Learning meets DVFS for Maximum Efficiency
- Host: GitHub
- URL: https://github.com/amirreza81/rl-dvfs-scheduling
- Owner: Amirreza81
- Created: 2025-01-29T19:17:56.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-02-12T20:19:17.000Z (3 months ago)
- Last Synced: 2025-02-12T21:27:41.649Z (3 months ago)
- Topics: dvfs, embedded-systems, energy-optimization, iot, multi-core, power-management, qos-optimization, real-time-scheduling, reinforcement-learning, rl, rl-based-scheduling, task-scheduling, tdp, thermal-design-power, thermal-management
- Language: Python
- Homepage:
- Size: 3.08 MB
- Stars: 3
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🚀 **Power Management & Task Scheduling in Real-Time Embedded Systems**
## **Project Overview**
This project focuses on **optimizing energy consumption** in embedded systems with limited resources, such as battery-powered devices. We utilize **reinforcement learning (RL)** and **dynamic voltage & frequency scaling (DVFS)** to efficiently manage **task scheduling**, ensuring system constraints are met.
---
## 🏗 **Project Phases**
### **🟢 Phase 1 - Offline Scheduling**
📌 Designed an **RL-based model** to optimize task scheduling on multi-core homogeneous systems.
📌 Applied **DVFS** to dynamically adjust **voltage & frequency** based on workload.
📌 Ensured **Thermal Design Power (TDP)** constraints were not violated.
📌 Used **Earliest Deadline First (EDF)** scheduling for real-time tasks.
### **🔵 Phase 2 - Online Scheduling**
📌 Integrated **soft aperiodic tasks** dynamically into the system.
📌 Developed an **adaptive task mapping strategy** 🗺 for real-time & aperiodic tasks.
📌 Optimized **Quality of Service (QoS)** while maintaining system constraints.
📌 Utilized **Slack Stealing Server** to efficiently handle task scheduling.
---
## 📊 **Results & Deliverables**
✅ Analysis of **task schedulability** under various workloads.
✅ Power consumption **trends for each core**.
✅ QoS **evaluation across different system states**.
✅ Final **implementation & report**.
---
## 🖼 **Visuals & Graphs**
---
📊 **CSV File:**
- [output_data.csv](./results/output6.csv)
| Task | Utilization | Period | WCET | Deadline | Arrival Time | Type | Frequency | Voltage |
|--------|-------------|--------|--------|----------|--------------|-------|-----------|---------|
| Task 0 | 0.1338 | 4 | 1.2845 | 4 | 0 | hard | 1.1 | 1.5 |
| Task 1 | 0.0285 | 5 | 3.4144 | 5 | 0 | hard | 0.75 | 0.9 |
| Task 2 | 0.0057 | 6 | 3.7932 | 6 | 0 | hard | 0.75 | 0.9 |
| Task 3 | 0.0147 | 1 | 0.6198 | 1 | 0 | hard | 1.1 | 1.5 |
| Task 4 | 0.0084 | 3 | 1.2931 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 5 | 0.0011 | 1 | 0.0147 | 1 | 0 | hard | 0.75 | 0.9 |
| Task 6 | 0.0747 | 5 | 3.0887 | 5 | 0 | hard | 0.75 | 0.9 |
| Task 7 | 0.0238 | 2 | 0.0702 | 2 | 0 | hard | 0.75 | 0.9 |
| Task 8 | 0.041 | 3 | 0.2194 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 9 | 0.0045 | 3 | 0.0239 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 10| 0.0401 | 3 | 0.7467 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 11| 0.0154 | 3 | 0.1211 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 12| 0.0018 | 6 | 0.0812 | 6 | 0 | hard | 0.75 | 0.9 |
| Task 13| 0.0592 | 3 | 1.7761 | 3 | 0 | hard | 0.75 | 0.9 |
| Task 14| 0.0083 | 2 | 0.1816 | 2 | 0 | hard | 0.75 | 0.9 |
| Task 15| 0.0642 | 1 | 0.2705 | 1 | 0 | hard | 0.75 | 0.9 |
| Task 16| 0.2006 | | 0.5258 | 10 | 9 | soft | 0.75 | 0.9 |
| Task 17| 0.253 | | 1.5719 | 9 | 0 | soft | 0.75 | 0.9 |
| Task 18| 0.2066 | | 5.4092 | 11 | 10 | soft | 0.75 | 0.9 |
| Task 19| 0.1854 | | 3.3038 | 8 | 3 | soft | 0.75 | 0.9 |
| Task 20| 0.1343 | | 3.8678 | 11 | 10 | soft | 0.75 | 0.9 |
---
## 📝 TXT Output Samples
Here are the log files generated for different core configurations and utilization levels:
### 🔹 **8 Cores**
- [scheduling_cores_8_utilization_0.25.txt](./results/scheduling_cores_8_utilization_0.25.txt)
- [scheduling_cores_8_utilization_0.5.txt](./results/scheduling_cores_8_utilization_0.5.txt)
- [scheduling_cores_8_utilization_0.75.txt](./results/scheduling_cores_8_utilization_0.75.txt)
- [scheduling_cores_8_utilization_1.0.txt](./results/scheduling_cores_8_utilization_1.0.txt)
### 🔹 **16 Cores**
- [scheduling_cores_16_utilization_0.25.txt](./results/scheduling_cores_16_utilization_0.25.txt)
- [scheduling_cores_16_utilization_0.5.txt](./results/scheduling_cores_16_utilization_0.5.txt)
- [scheduling_cores_16_utilization_0.75.txt](./results/scheduling_cores_16_utilization_0.75.txt)
- [scheduling_cores_16_utilization_1.0.txt](./results/scheduling_cores_16_utilization_1.0.txt)
### 🔹 **32 Cores**
- [scheduling_cores_32_utilization_0.25.txt](./results/scheduling_cores_32_utilization_0.25.txt)
- [scheduling_cores_32_utilization_0.5.txt](./results/scheduling_cores_32_utilization_0.5.txt)
- [scheduling_cores_32_utilization_0.75.txt](./results/scheduling_cores_32_utilization_0.75.txt)
- [scheduling_cores_32_utilization_1.0.txt](./results/scheduling_cores_32_utilization_1.0.txt)
---
## 👥 **Members**
We were a team of two members consisting of the following individuals:
- [AmirReza Azari](https://github.com/Amirreza81)
- [Bozorgmehr Zia](https://github.com/BozorgmehrZia)