Research Assistant (Drone Swarm in Cluttered Environments)

NATIONAL UNIVERSITY OF SINGAPORE; Singapore; On-site; SGD 60,000 - 90,000; Full time

Job summary

A leading educational institution in Singapore is seeking a candidate to develop deep multi-agent reinforcement learning methods for drone collaboration in urban environments. Responsibilities include investigating traditional and AI strategies for effective swarm planning while handling constraints like limited computation. Candidates should have excellent coding abilities in Python and PyTorch, experience with drone deep learning applications, and strong communication skills. This role offers a unique opportunity to contribute to cutting-edge AI research and development.

Qualifications

• Proficient in distributed deep reinforcement learning and Transformers.
• Ability to summarize and review literature effectively.
• Experience with simulation environments like Isaac Lab and ROS Gazebo.

Responsibilities

• Develop multi-agent reinforcement learning approaches for drones.
• Investigate conventional and AI-based methods for swarm planning strategies.
• Create simulation environment for training AI policies and deploy controllers on hardware.

Skills

Excellent coding skills in python with pytorch

Literature review/summarizing skills

Simulations abilities (e.g., Isaac Lab, ROS Gazebo)

Experience with implementation of deep learning model on aerial robots

Experience publishing papers

Good writing/spoken communication skills

Job description

Interested applicants are invited to apply directly at the NUS Career Portal

Your application will be processed only if you apply via NUS Career Portal

We regret that only shortlisted candidates will be notified.

Job Description

This project focuses on developing deep multi-agent reinforcement learning approaches for a team of multi-rotor drones to collaboratively search for targets in low-rise urban environments. These environments are usually characterized by cluttered structures, GNSS-denied conditions, low-light scenarios, and real-world constraints such as limited onboard computation and communication bandwidth. The candidate will investigate both conventional pipelines, including mapping and motion planning, and recent AI-based methods to design robust, scalable, and efficient swarm planning strategies. Each drone will be equipped with LiDAR and camera sensors for mapping and navigation, potentially enhanced by semantic perception to improve decision-making and coordination. The candidate will develop both simulation environment for training AI policy and deploy the controller on hardware.

Qualifications

• Excellent coding skills in python with pytorch (distributed deep reinforcement learning, Transformers, etc.)
• Literature review/summarizing skills
• Simulations abilities (e.g., Isaac Lab, ROS Gazebo)
• Experience with implementation of deep learning model on aerial robots is preferred
• Experience publishing papers, and supervising undergraduate/master’s students
• Good writing/spoken communication skills