Aryan
Recent News
New Paper Submitted: My latest research SymAnt International Journal of Robotics Research.
Hello! I am an MS student in Robotics at The Robotics Institute, Carnegie Mellon University, advised by Prof. Katia Sycara.
I have a passion for exploring how human-level symbolic common sense can be integrated into AI agents, particularly in computer vision algorithms. In pursuit of this passion, I aim to develop human-inspired computer vision systems, focusing on Generative AI and perception for autonomous vehicles. As part of my Master’s thesis, I am working on embedding symbolic knowledge through scene and knowledge graphs to enable common sense reasoning for tasks such as compound scene prediction and action anticipation.
I received my Bachelor's degree in Electronics and Communication from Delhi Technological University. During my undergraduate studies, I led the Unmanned Ground Vehicle Unmaneed Ground Vehicle team at DTU, where we developed an autonomous vehicle from scratch. I was also a 2022 CMU Robotics Institute Summer Scholar, working with Prof. Sebastian Scherer and Prof. Chen Wang.
Active Projects
To know more about my current and past experiences, visit the Experience page.
Featured Publications
Professional Experience
Neuro-Symbolic Models for Interpretable Scene and Behaviour Understanding
Advanced Agents Robotics Technology Lab, CMU - Research Assistant
May 2023 | GPA: 9.5/10
Publications: Submission - ICCV 2025 Code Paper, Submission - IROS 2025 Code Paper, IEEE SPIN 2023 Code Paper
- Developed sample-efficient diffusion models to anticipate long-term human actions from short video observations.
- Combined scene and knowledge graphs to provide interpretable, contextually rich priors to neural networks.
- Improved state-of-the-art anticipation accuracy by 4-5% while reducing model size by 80x and training time by 3x.
Robotics Institute Summer Scholarship, CMU
Developed AirLoc, a simple yet effective indoor relocalization approach that utilizes both geometric and semantic information of a scene by using objects to perform relocalization.
- Outperformed state-of-the-art methods in room-level relocalization.
- Supervisor: Prof. Sebastian Scherer (Personal Webpage)
Carnegie Mellon University - Pypose Project
Worked on developing a first-of-its-kind differentiable IMU preintegrator and added bias correction and covariance propagation to the base IMU preintegrator for PyPose, a library connecting classical learning with modern methods.
- Supervisor: Prof. Sebastian Scherer (Personal Webpage)
International Institute of Information Technology, Hyderabad
Worked on a novel approach for place-recognition using Scene-Graphs and deep-learning techniques for room-level and metric-level localization.
- Extracted baseline accuracy from existing techniques like NetVlad, Scan Matching, and DeLG.
- Supervisor: Dr. K. Madhava Krishna (Personal Webpage)
Samsung Digital Academy Research Lab, DTU
Developed a stress prediction system based on Electrocardiogram (ECG) and Heart Rate Variability (HRV) data, determining optimal HRV features for cumulative stress monitoring.
- Tested various machine learning and deep learning algorithms to improve accuracy over baselines.
- Supervisor: Dr. Divyashikha Sethia (Personal Webpage)
2024
2023
2022
Self-Supervised Flare Removal from Images
(August 2021 - Present, DTU, New Delhi, India)
Developed a framework for training a flare removal network using random flare and flare-free images extracted from the internet, eliminating the need for labels.
- Worked on synthetic data generation required to train the algorithm.
Legal Document Summarization with Fine-tuned LLM and RAG Pipeline Code
(Jan 2024 - May 2024, Carnegie Mellon University, Pittsburgh)
Developed a robust pipeline for summarizing lengthy Indian legal documents by fine-tuning a large language model and integrating a retrieval-augmented generation system.
- Used LoRA to adapt Llama-2-7b for summarizing lengthy Indian legal documents with domain-specific terms.
- Deployed a RAG pipeline using FAISS and TF-IDF for efficient document retrieval and scalable summarization.
Griffin - Unmanned Ground Vehicle
(August 2019 - August 2021, DTU, New Delhi, India)
Developed an autonomous ground vehicle to represent Delhi Technological University in the Intelligent Ground Vehicle Competition at Oakland University, Rochester, Michigan.
- Engineered the navigation stack of a self-driving rover using Robot Operating System (ROS) with lane detection and obstacle avoidance.
- Utilized ROS packages such as Slam_gmapping, Move_base, and Costmap for synchronization with real-world environments.
- Implemented sensor fusion with IMU, GPS, Lidar, odometers, micro-controllers, and actuators.
- Supervisor: Dr. S. Indu (Personal Webpage)
Study of Neural Network Implementation on FPGAs
(Jan 2021 - May 2021, DTU, New Delhi, India)
Compared different methods to implement neural networks on FPGAs and determined the best approach for early prototyping using high-level languages like Python.
Micro-Mouse Challenge
(December 2020, IIT Bombay TechFest)
Participated in the Micro-Mouse simulation challenge at IIT Bombay's TechFest, solving a maze using a simulation in Gazebo.
- Implemented a backtracking algorithm for maze exploration and created a map using the Slam_gmapping package in ROS.
- Project Documentation (Drive Link)
Line Follower Minidrone
(April 2021 - June 2021, MATLAB Minidrone Competition)
Developed a line-follower algorithm based on model-based design using MATLAB and Simulink for a Parrot minidrone.
Localization Package for Unmanned Aerial Vehicle
(April 2021 - July 2021, India)
Developed a localization package for UAVs in dim-light and GPS-denied environments with Octobotics Tech Private Limited.
- Developed using Robot Operating System (ROS).
Ministry of Electronics and Information Technology Sponsored R&D Project
(Jan 2021 - July 2021, India)
Worked on developing and deploying motor controllers for low and medium-power electric vehicles.
- Reduced the average temperature of MOSFETs by 10 degrees through optimal heat-sink design.
JN Tata Endowment for Higher Education
Top 0.1% in a pool of 10,000+ applicants
Mar, 2023Robotics Institute Summer Scholarship
One of the 4 students from India selected for the prestigious research internship at Carnegie Mellon University
Mar, 2022Intelligent Ground Vehicle Competition
Secured 1st position in the Cybersecurity Challenge for the IGVC competition held at Oakland University in Rochester, Michigan
Mar, 2022DAAD WISE Internship
One of the 80 students from India selected for research in Germany
Mar, 2022Mitacs Globalink Internship
One of the 117 students from India selected for an internship in Canada (M.S.)
Mar, 2022MATLAB Minidrone Competition
Qualified for the International Finale in MATLAB Minidrone Competition. Our team was one of the top 8 teams from Europe, Middle East, and Asia for Line Following Minidrone.
Oct, 2021