Hello, I'm Jasmine!

I'm a first year PhD student at the Massachusetts Institute of Technology, USA.

Jasmine Jerry Aloor

First Year Graduate Student,(MIT)

Major: Aeronautical And Astronatical Engineering (MIT AeroAstro)

Robotics has always captivated me. Autonomous systems are at the forefront, but appropriate optimization and intelligent consideration of the tradeoffs involved can help these approaches reach their full potential. I am keen on exploring motion planning and control for dynamic multi-agent systems, with a special emphasis on safety, robustness, and human interactivity. I have worked on imitation learning for social navigation, control barrier functions, and coverage path planning in my past research experiences combining Control Theory, Machine Learning and Computer Vision.

I am pursuing graduate studies from Fall 2022 at DINaMo Group, MIT with Professor Hamsa Balakrishnan to strengthen my knowledge further and contribute to the research community. I aim to use a design-based approach backed by strong fundamentals, to help create a breakthrough in my field.

Medium Account My LinkedIn Profile Github Account

Get to know about my skills below!

Robotics

Click to expand

From my freshman year, I began dabbling in the various fields of robotics beginning with embedded systems and controls, to computer vision and deep learning

Soccer Playing Ground Robots (RoboSoccer)
Aerial Robots
Space Robotics

Aerospace Engineering

Click to expand

I'm fascinated by all things aerospace and am particularly interested in

Aerospace Control Systems
Aircraft Design
Aeromodelling

Artificial Intelligence

Click to expand

Robotics is something I am passionate about. I am confident and have applied these technologies for robots and aircraft systems

Computer Vision
Deep Learning
Algorithms

This section showcases some of my best and latest work I have created with links to the projects and a brief description.

Here are some of my research experiences

Socially Aware Imitation Learning in Multi-Aircraft Shared Airspace

Developing safe and socially aware motion planning method using imitation learning for general aviation (GA) aircraft in multi-aircraft shared airspace. Incorporating safety through formal logic methods.
Robotics Institute Summer Scholar RISS Program, AirLab, Carnegie Mellon University

Social Navigation STL Multi Agent System
Preprint Working Paper Preliminary Research Showcase

Multi UAV Trajectory Tracking and Obstacle Avoidance

Developed and simulated a pair of UAVs to track a moving target. Implemented Barrier Lyapunov Functions to maintain desired formation. Internship at the GCDS Lab, IISc.

ROS Gazebo Obstacle Avoidance Barrier Functions
Paper Video

Deep Learning based Manipulator Arm Inverse Dynamics

Data driven approach to learn inverse dynamics of a robotic manipulator arm. Developed a setup on ROS to learn model dynamics when model information is not provided.
Worked remotely at McGill University as a part of MITACS Globalink Research Internship.

Deep Learning Manipulation Inverse Dynamics

Agri-Drone Pest Detection and Path Planning

Coverage path planning for agricultural field. Optimal trajectory, safe and robust control using Control Barrier Functions. Bachelor's Thesis Project

Deep Learning Agri Drone Path Planning

FLYing maniPULATOR: Fully Actuated Aerial Manipulator Project

Analysed the omnidirectional controllability, studied nonlinear control theory, and dynamics of existing system for fault-tolerant control. Transitioned project software stack ROS1 to ROS2 and optimized the code.
Worked remotely with TU Dresden as a part of DAAD WISE Scholarship 2020.

Nonlinear Controllability ROS2

Below are some of my Robotics Projects!

RoboSoccer Small Sized League

Among top 25 teams in the world that qualified for SSL RoboCup 2019, a robot soccer competition. Worked on motor control and embedded systems. Results to be published soon.

RoboCup KRSSG Embedded
Team's GitHub

Crusade Robotix

IIT Kharagpur's tech fest Kshitij 2019 Image processing Robot based on Raspberry Pi.

Tech Fest Image Processing RPi
GitHub

Fuzzy Logic Control for PID

Developed and tested an optimized Fuzzy control code for motor velocities. Applied on the RoboSoccer 2020 robots. Showed robustness even when disturbances were applied.

Fuzzy Velocity Control
GitHub

Here are some of the aerospace and controls projects that I have done!

Box Wing MAV

Built, tested and studied properties for an improvised box wing MAV. Presented at the National Conference of Wind Tunnel Testing, Kanpur, India (2020)

MAV Box Wing Aerodynamics
Preprint Video Conference Paper

Range optimization of a guided projectile

Using a Homotopy method, determine the maximum range for a guided projectile based on optimal controls.

Trajectory control Homotopy Optimize
GitHub

National Aerospace Conceptual Design Competition 2, 3, 4

Won 3rd Place NACDeC-2: Designed a solar powered and fuel cell combination twin propeller high aspect ratio wing aircraft. Competition by the Aeronautical Society of India (AeSI). Led the Department Team for NACDeC-4.

Aircraft Design Conceptual Design

Some additional projects are listed below

Panoramic Image Generation

Automated tool that performed multiple image stitching followed by blending. Semester Project, Image Processing, Autumn 2018.

Image Processing
GitHub

MathWorks Minidrone Competition 2021

Led the IITKGP Aerospace Team for MathWorks Napoli Minidrone Competition 2021, to develop algorithm for a line follwing drone on Simulink. Event cancelled after submission round.

Coming soon! Robotics and AI

AUVSI Student Unmanned Aerial Systems (SUAS)

UAV club team to participate in the Association for Unmanned Vehicle Systems International (AUVSI) Student Unmanned Aerial Systems (SUAS) Competition 2021. Event delayed due to COVID-19

UAV Team Head

End-to-end Goal Conditioned Social Navigation using Signal Temporal Logic

Robotics Institute Summer Scholar Program

With an anticipated increase in aerial traffic in the near future in the form of crewed air-taxis, aircraft, and autonomous aerial systems, achieving safe human-operated and independent vehicle coordination in shared airspace is crucial. Current studies have investigated pedestrian and ground vehicle trajectories in multiagent environments in great detail. However, these studies have not yet analyzed aerial trajectories. There is a need to have safely separated aircraft in the vicinity of airports that behave as required when entering and leaving. This work takes a step forward to achieve safe manned-unmanned vehicle operations while learning from each other. I developed a motion planning algorithm using methods of behavior cloning. I used the current state, past trajectory, and intent to determine the subsequent actions with a discriminator reward based method. Using real-world aircraft flight data from the TrajAir dataset, augmented with the agent’s goal to generate optimal trajectories as future actions. Our system incorporates continuous learning allowing autonomous aircraft to integrate into regular manned traffic flow safely.

The key contributions in my work include:

Usage of beam search that couples motion prediction with planning. Beam search rolls out our policy and selects the best trajectory plan based on variable metrics.
Utilization of Signal Temporal Logic as the formal logic mechanism to ensure contstaints are modelled accurately and satified.
Implementing multi-agent beam search method for motion planning of multiple dynamic agents.

This image shows the results of the single agent beam search. Beam Tree generated based on two metrics: (a), (b): trajectory goal has least Euclidean distance from original goal location; (c), (d): trajectory goal to match original goal-vector

We encode the metrics of reaching desired goal and selecting trajectories away from a threshold distance from other agents for collision avoidance (hard constraint) using signal temporal logic (STL).