Enhancing Autonomous Vehicle Performance Through Adaptive FPGA Acceleration and Reinforcement Learning for Dynamic Environments

Project ID: AIRG/2023/EEE/05/01

Funding Source: AUST Internal Research Grant

Project Awarded: 1^st April, 2024

 

Principal Investigator:

Dr. Fakir Sharif Hossain

Associate Professor, EEE Department, AUST

 

Co-principal Investigator/s:

Ashek Seum

Lecturer, CSE Department, AUST

 

Md. Reasad Zaman Chowdhury

Lecturer, CSE Department, AUST

 

Project Duration: 2 years

Budget approved: 10,00,000/-

 

Expected Research Outcomes:

The expected outcomes of this research project include improved safety, adaptability, efficiency and user experience in AV systems.

• The integration of sensor data and reinforcement learning within the FPGA is expected to lead to improved real-time decision-making for AVs.
• The dynamic selection and configuration of FPGA hardware accelerators are anticipated to optimize resource utilization and improve task-specific performance.
• The system’s ability to respond to changing environmental conditions such as traffic congestion and weather is expected to enhance the vehicle’s adaptability and safety.
• Over time, the reinforcement learning agent should continuously improve its decision-making abilities, leading to more efficient and safer autonomous driving.
• By dynamically optimizing hardware resources and decision-making, the method should lead to increased operational efficiency for AVs.