Hello,
My name is Varun Murali.
I am currently a postdoc at the University of Pennsylvania working on hierarchical semantic mapping, planning and tasking for robots in the wild.
Recent News:
• September '23: I defended my thesis titled "Perception-aware planing for differentially flat robots"!
• November '22: I am giving the robotics seminar at Cornell about our work on "Perception-aware planning for fast and agile robots"
• June '20: Our work on "Joint Feature Selection and Time Optimal Path Parametrization for High Speed Vision-Aided Navigation" is accepted to appear at IROS
• January '22: Our work on "A Planted Clique Perspective on Hypothesis Pruning." is accepted to appear in RA-L!
• December '21: Completed the Kaufman Teaching Certificate at MIT.
• May '21: Completed my thesis proposal.
• January '21: Our work on "6D Object Pose Estimation with Pairwise Compatible Geometric Features" is accepted to appear at ICRA!
• August '20: Our work on FlightGoggles is featured on MIT spectrum!
Decision Making Under Uncertainty
Most autonomous vehicles, especially high-performance aircraft, must perceive their environment very rapidly using limited computing and sensing resources. In many high-performance applications, slight deviations in the trajectory of the vehicle may enable substantial increase in perception capabilities. Therefore, the trajectory of the vehicle must be chosen carefully to optimize, not only trajectory objectives, but also perception objectives.
Photorealistic Simulation for Robotics
FlightGoggles is envisioned to be development environment that allows the design, implementation, testing and validation of autonomous super-vehicles. FlightGoggles currently provides exteroceptive sensor simulation based on the Unity3D engine as well as vehicle dynamics and inertial sensor simulation capabilities. We aim to train algorithms in this photogrammetric environment and transfer them directly to real world applications.
6D Object Pose Estimation
Pose estimation is critical in many robotics applications, particularly to enable autonomous vehicles to perceive other vehicles around them. We design algorithms that allow 6D pose estimation in challenging scenarios, including heavy occlusion, while under computational constraints.