Design and optimization of scheduling algorithms for efficient resources utilization on digital microfluidic biochips

نوع: Type: Thesis

مقطع: Segment: masters

عنوان: Title: Design and optimization of scheduling algorithms for efficient resources utilization on digital microfluidic biochips

ارائه دهنده: Provider: Atefeh Nourouzi

اساتید راهنما: Supervisors: Dr. Mahdi Abbasi – Dr. Hatam Abdoli

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Dr. Abbas Ramezani – Dr. Shakoor vakilian

زمان و تاریخ ارائه: Time and date of presentation: 2025

مکان ارائه: Place of presentation: seminar

چکیده: Abstract: Abstract: In recent years, Digital Microfluidic Biochips (DMFBs) have gained significant attention for various safety-critical and biomedical applications such as clinical and point-of-care diagnostics, drug discovery, biochemical analysis, and other related fields. These chips have effectively miniaturized traditional laboratory spaces, making biochemical experiments more compact, efficient, and user-friendly compared to conventional methods. For complex assays, routing and transporting droplets while satisfying multiple physical and operational constraints is a challenging task. Since medical applications directly impact human health, accuracy in experimentation and the speed of disease detection are of utmost importance. However, DMFBs face various faults and challenges that can compromise their reliability due to their limited fault-tolerance capabilities. Such failures may cause droplet routing processes to halt or malfunction. These faults are generally classified into known and unknown categories. Known faults are typically detected before routing through sensors, while unknown faults—such as the sudden breakdown of electrodes—may occur during operation and disrupt system performance. Therefore, the proposed algorithm must be capable of effectively detecting and managing both known and unknown faults, ensuring uninterrupted and successful routing. In this study, aiming to improve performance and reliability in DMFB-based systems, intelligent scheduling approaches have been explored. Unlike classical scheduling algorithms that operate statically based on predefined rules, this work employs Reinforcement Learning (RL) algorithms to learn dynamic and optimal scheduling policies. After analyzing various RL methods, the Advantage Actor-Critic (A2C) algorithm was selected and implemented as the final approach. By continuously interacting with the simulated environment, A2C optimizes decisions related to operation scheduling, resource allocation, and droplet routing. Experimental results and comparisons with baseline algorithms demonstrate that A2C achieves superior stability, exploration efficiency, computational complexity, and average reward, indicating its strong potential for application in complex bio-assay environments.