Research Fellow - Multi-Agent RL for Autonomous Drone Swarm

About the role

The School of Mechanical & Aerospace Engineering (MAE) at NTU is looking for a Research Fellow to develop learning-based algorithms for cooperative target tracking in complex environments. The role focuses on multi-agent reinforcement learning, decentralized target assignment, occlusion-aware decision-making and communication-resilient coordination for autonomous drone teams operating in urban and cluttered environments.

Key responsibilities

• Develop learning-based frameworks for cooperative multi-agent robotic systems in complex environments.
• Formulate multi-agent decision-making problems, including state and action representation, reward design, task allocation and decentralized policy learning.
• Develop reinforcement learning and multi-agent reinforcement learning algorithms for autonomous coordination and target-following under uncertainty, partial observability and dynamic conditions.
• Develop perception-aware decision-making methods and integrate perception, decision-making and control modules within a simulation-based validation framework.
• Design and run simulation experiments to evaluate performance, robustness, scalability and generalization.
• Work with PhD students, research engineers and collaborators on system integration, testing and demonstration.
• Prepare technical reports, publications, presentations and project deliverables.

Required profile

• PhD in Robotics, Aerospace, Mechanical, Electrical and Electronic Engineering, Computer Science, Artificial Intelligence or a closely related discipline.
• Strong research background in multi-agent reinforcement learning, multi-robot systems, autonomous systems or learning-based navigation.
• A strong publication record in relevant journals or conferences is an advantage.
• Ability to mentor junior researchers, PhD students or research engineers, and willingness to support project reporting and milestone reviews.

Required skills

• Python and deep learning frameworks such as PyTorch or TensorFlow.
• Reinforcement learning and multi-agent reinforcement learning algorithms.
• Simulation environments such as Unity, ROS/ROS2, Gazebo or AirSim.
• Multi-agent coordination, decentralized control, target assignment or swarm robotics.
• Graph neural networks, attention mechanisms, computer vision, sensor fusion or multi-object tracking are advantageous.
• Experience training RL policies in simulation and with drone swarms or UAV coordination; prior real-world UAV experiments are an advantage but not mandatory.