Research Topics & Areas
Explore our active research areas and find where your interests align with ARENA Lab.
Robot Navigation
Autonomous navigation systems for mobile robots in complex, dynamic environments. Classical planning to learning-based approaches.
Social Navigation
Robots navigating among humans in a socially compliant manner. Understanding human motion patterns and social norms.
Simulation & Digital Twins
High-fidelity simulation environments for training and evaluating robot behaviors. Bridging the sim-to-real gap.
Generative World Models
Generative models to learn environment dynamics for improved robot planning in unseen scenarios.
VLA/VLM for Robotics
Vision-Language-Action and Vision-Language Models enabling robots to follow natural language instructions.
Human-Robot Interaction
Intuitive interfaces and behaviors for robots working alongside humans. User studies, trust, and safety.
Benchmarking & Evaluation
Standardized benchmarks and metrics for evaluating robot navigation and social compliance.
Sim-to-Real Transfer
Transferring policies trained in simulation to real robots. Domain adaptation, randomization, robustness.
ARENA Platform
Contributing to the open-source ARENA simulation and benchmarking platform. Software development and community building.
Embodied AI
Building AI systems that interact with the physical world through robotic embodiments. Combining perception, planning, and action.
Active Projects
Quadrupeds in Crowded Environments
Integration of quadrupeds into ArenaNav. Goal: deploy real quadruped robots at NUS campus and food centers!
Human Detection and Tracking (RGB-D)
Computer vision for human detection, tracking, action and social state detection communicated via ROS.
Arena-Web v3
Third version of our web-based benchmarking platform. Contribute new functionalities and participate in a paper.
ArenaNav v4 – Core Platform
Fourth version of our core development and benchmarking platform. Always looking for students to contribute and participate in papers.
Photorealistic Simulation
Integration of ISAAC Gym simulation into ArenaNav and development of multimodal training approaches for navigation in crowds.
Robot Behavior Adaptation via Semantic Mapping
Navigation and behavior adaption for mobile robots based on semantic mapping and social interactions.
Multi Agent Reinforcement Learning
Multi-Agent-RL approaches for robots (homogeneous/heterogeneous) for tasks such as rescue or evacuation.
LLM-based World Generation
Leveraging LLMs to dynamically generate millions of randomized worlds for training and benchmarking.