Siddarth Narasimhan

I'm an MASc student at the University of Toronto Robotics Institute, advised by Dr. Goldie Nejat, where I work on building novel navigation and control policies for robots using diffusion and large language models.

I completed my BASc in Engineering Science, Robotics Option and AI Minor at the University of Toronto with Honours. Outside of research, you will find me either pursuing the National Master title, performing in concerts on the mridhangam, or playing competitive table tennis.

Last Updated: 09/24

Email  /  Resume  /  LinkedIn  /  Scholar  /  GitHub

Publications

We introduce OLiVia-Nav, an online lifelong vision language architecture for mobile robot social navigation. By leveraging large vision-language models (VLMs) and a novel distillation process called SC-CLIP, OLiVia-Nav efficiently encodes social and environmental contexts, adapting to dynamic human environments.

We present a novel robot exploration map prediction method called Confidence-Aware Contrastive Conditional Consistency Model (4CNet), to predict (foresee) unknown spatial configurations in unknown unstructured multi- robot environments with irregularly shaped obstacles.

Experience

Documented performance of ASICs using metrics such as power efficiency, over current protection, dynamic response and power up. Inspected schematics and layouts to rework damaged boards and recommend improvements. Designed a robust Remote Management platform using a Raspberry Pi to automate the generation and sending of I2C/SMBUS/ MCTP/PLDM packets, and decode and debug live traffic through a custom-built GUI. Received Spotlight Award for novel contributions and excellent performance as a co-op student.

Developed an intelligent provincial transportation system for highway analytics by leveraging GPS data and machine learning. Used by the Ministry of Transportation to obtain live traffic volume estimates on GTA highways. Designed a novel timestamp detection and recognition algorithm to locate and convert timestamps found in highway video feed to text.

Projects

An object-oriented C++ implementation of popular variants of the rapidly exploring random trees algorithm, including RRT, RRT*, Anytime RRT and Informed RRT*.

The goal of RobotVision is to simulate control, path-planning, localization and mapping of a non-holonomic robot in a 2D world, while also providing a comprehensive mathematical formulation of each method. Implementation includes PID Control, lead-lag control, Dubin's path calculator, Extended Kalman Filter, and SLAM.

For our capstone project, we built a drone that is capable of stable hover, waypoint navigation and obstacle avoidance, for use in a nuclear plant facility. Our team won the award for smoothest flight.

Scholarships

Teachings

Extracurricular Achievements

Thank you to Jon Barron.