Lakshita Dodeja

Hi! I am Lakshita, a second year PhD student at Brown University, advised by Prof. Stefanie Tellex in the Humans to Robots Lab. I'm broadly interested in Robot Learning, Human-Robot Interaction and Natural Language Processing. More recently, I have been working on adapting robot manipulation policies, based on foundation models, behavior cloning and diffusion policies to the changes in environment using residual policy learning. I also completed my Master's in Computer Science at Georgia Tech with a specialisation in Computational Perception and Robotics where I worked in the CORE Robotics Lab advised by Prof. Matthew Gombolay. During my Master’s I worked on Human-Robot Interaction specifically on distilling strategic intent of users from unstructured natural language commands and designing user centric strategy recommendation systems for Human-AI tasks. Previously, I also worked as Software Development Engineer at Amazon, India in the Prime Verification Team. I did my undergrad studies from NIT Kurukshetra, India where I worked with Dr Mayank Dave on Digital Image Watermarking and Wireless Networks.

Email  /  CV  /  Linkedin  /  Google Scholar  /  Github  /  Twitter

Artificial Intelligence is being employed by humans to collaboratively solve complicated tasks for search and rescue, manufacturing, etc. Efficient teamwork can be achieved by understanding user preferences and recommending different strategies for solving the particular task to humans. In this paper, we seek to understand the important factors to consider while designing user-centric strategy recommendation systems for decision-making. We conducted a human-subjects experiment (n=60) for measuring the preferences of users with different personality types towards different strategy recommendation systems.

Many real-world tasks involve a mixed-initiative setup, wherein humans and AI systems collaboratively perform a task. While significant work has been conducted towards enabling humans to specify, through language, exactly how an agent should complete a task (i.e., low-level specification), prior work lacks on interpreting the high-level strategic intent of the human commanders. In this paper, we build a computational interface capable of translating unstructured language strategies into actionable intent in the form of goals and constraints.

With emergence of new technologies it is now easier to communicate through multimedia like image, audio, video and text. But at the same time the problem of unauthorized access and copyright protection has also emerged. In order to handle these problems digital image watermarking is one of the best technique. In this paper we present an optimized color image watermarking technique to protect an image data from any unauthorized access. The technique presented in the paper uses a combination of Discrete Stationary Wavelet Transform (DSWT), Singular Value Decomposition (SVD), Discrete Cosine Transform (DCT) and Arnold Transform.