PhD Thesis Proposal
Carnegie Mellon University
Structured Learning for Robust Robot Manipulation
Abstract: Robust and generalizable robots that can autonomously manipulate objects in semi-structured environments can bring material benefits to society. Data-driven learning approaches are crucial for enabling such systems by identifying and exploiting patterns in semi-structured environments, allowing robots to adapt to novel scenarios with minimal human supervision. However, despite significant prior work in learning for [...]
Simulation-based Planning for Pick-and-Place in Heavy Clutter using Non-prehensile Manipulation
Abstract: Robot manipulation in domestic households, industrial manufacturing and warehouses might require contact-rich interactions with objects in the environment. For pick-and-place style grasping tasks in cluttered scenes, it can be more economical for the robot to rely on non-prehensile actions vis-à-vis deliberate prehensile rearrangement. Non-prehensile actions also let the robot manipulate large and bulky objects [...]
Carnegie Mellon University
Learning with Diverse Forms of Imperfect and Indirect Supervision
Abstract: High capacity Machine Learning (ML) models trained on large, annotated datasets have driven impressive advances in several fields including natural language processing and computer vision, in turn leading to impactful applications of ML in areas such as healthcare, e-commerce, and predictive maintenance. However, obtaining annotated datasets at the scale required for training such models [...]
Carnegie Mellon University
3D Representation Learning for Perception and Prediction: A Modular Yet Highly Integrated Approach
Abstract: Modularized and cascaded autonomy stacks (object detection, then tracking and then trajectory prediction) have been widely adopted in many autonomous systems such as self-driving cars due to its interpretability. In this talk, I advocate the use of such a modular approach but improve its accuracy and robustness by developing different 3D representations for each [...]
Carnegie Mellon University
Understanding Unbalanced Datasets Through Simple Models and Dataset Exploration
Abstract: Computer vision models have proven to be tremendously capable of recognizing and detecting several classes and objects. They succeed in classes widely ranging in type and scale from humans to cans to pens. However, the best performing classes have abundant examples in large-scale datasets today. In unbalanced datasets, where some categories are seen in [...]
Carnegie Mellon University
Self-Supervising Occlusions for Vision
Abstract: Virtually every scene has occlusions. Even a scene with a single object exhibits self-occlusions - a camera can only view one side of an object (left or right, front or back), or part of the object is outside the field of view. More complex occlusions occur when one or more objects block part(s) of [...]
Carnegie Mellon University
Massively Parallelized Lazy Planning Algorithms
Abstract: Search-based planning algorithms enable autonomous agents like robots to come up with well-reasoned long horizon plans to achieve a given task objective. They do so by optimizing a task-specific cost function while respecting the constraints on either the agent (e.g. motion constraints) or the environment (e.g. obstacles). In robotics, such as in motion planning [...]
Run-Time Optimization in the Deep Learning Age
Abstract: In a recovery task one seeks to obtain an estimate of an unknown signal from a set of incomplete measurements. These problems arise in a number of computer vision applications, from image based tasks such as super-resolution and in-painting to 3D reconstruction tasks such as Non-Rigid Structure from Motion and scene flow estimation. Early [...]
Carnegie Mellon University
System Identification and Control of Multiagent Systems Through Interactions
Abstract: This thesis investigates the problem of identifying dynamics models of individual agents of a multiagent system (MAS) and exploiting these models to shape their behavior using robots extrinsic to the MAS. While task-based control of a MAS using onboard controllers of its agents is well studied, we investigate (a) how easy it is for [...]
Carnegie Mellon University
Driving Reconfigurable Unmanned Vehicle Design for Mobility Performance
Abstract: Unmanned ground vehicles are being deployed in increasingly diverse and complex environments. Advances in the field of robotics, including perception technology, computing power, and machine learning, have brought robots from the lab to the real world. Remote and autonomous vehicles are now used to explore volcanoes, caves, pipes, war zones, disaster sites, and even [...]
Towards Large-scale and Long-term Neural Map Representations
Abstract: We address the problem of large-scale and long-term neural map representations. Maps, as our prior understanding toward the environment, provide valuable information for modern robotics applications such as autonomous driving and AR/VR. The size of maps largely affects the end task performance: usually a more detailed map can support better performance, but would cost [...]
Manipulating Objects with Challenging Visual and Geometric Properties
Abstract: Object manipulation is a well-studied domain in robotics, yet manipulation remains difficult for objects with visually and geometrically challenging properties. Visually challenging properties, such as transparency and specularity, break assumptions of Lambertian reflectance that existing methods rely on for grasp estimation. On the other hand, deformable objects such as cloth pose both visual and [...]
Large Scale Dense 3D Reconstruction via Sparse Representations
Abstract: Scene reconstruction systems take in (3D) videos as input, and output 3D models with associated poses for inputs. With the demand of 3D content generation, the technique has been drastically evolving in recent years. For professionals equipped with depth sensors, efficient dense reconstruction systems have become available to efficiently recover scene geometry. For ordinary [...]
Teaching Agent Reward Functions via Demonstrations for Human Inverse Reinforcement Learning
Abstract: For intelligent agents (e.g. robots) to be seamlessly integrated into human society, humans must be able to understand their decision making. For example, the decision making of autonomous cars must be clear to the engineers certifying their safety, passengers riding them, and nearby drivers negotiating the road simultaneously. As an agent's decision making can [...]
Policy Decomposition: Approximate Optimal Control with Suboptimality Estimates
Abstract: Optimal Control is a formulation for designing controllers for dynamical systems by posing it as an optimization problem, whereby the desired long-term behavior of the system is expressed using a cost function. The objective is to compute a policy, i.e. a mapping from the state of the system to its control inputs, that minimizes [...]
Audience-Aware Legibility for Social Navigation
Abstract: Robots often need to communicate their goals to humans when navigating in a shared space to assist observers in anticipating the robot’s future actions. These human observers are often scattered throughout the environment, and each observer only has a partial view of the robot and its movements. A path that non-verbally communicates with multiple [...]
On Sample-Efficient Reinforcement Learning for Nuclear Fusion
Abstract: In many practical applications of reinforcement learning (RL), it is expensive to observe state transitions from the environment. For example, in the problem of plasma control for nuclear fusion, determining the next state for a given state-action pair requires querying an expensive transition function which can lead to many hours of computer simulation or [...]
Carnegie Mellon University
Towards reconstructing non-rigidity from single camera
Abstract: In this proposal, we study how to infer 3D from images captured by a single camera, without assuming the target scenes / objects being static. The non-static setting makes our problem ill-posed and challenging to solve, but is vital in practical applications where target-of-interest is non-static. To solve ill-posed problems, the current trend in [...]
Efficient 3D Representations: Algebraic Surfaces for Differentiable Rendering
Abstract: In this proposal, we show how some classic computer vision tasks can robustly be solved via optimization techniques by using an object representation that is compact and interpretable. Specifically, we explore the applications and benefits of representing 3D objects with an analytical, algebraic function by building an approximate, ray-based differentiable renderer. Our approximate formulation [...]
Continual Robot Learning: Benchmarks and Modular Methods
Zoom Meeting Passcode: 841755 Abstract: The earliest reinforcement learning models were designed to learn one task, specified up-front. However, an agent operating freely in the real world will not in general be granted this luxury, as the demands placed on the agent may change as environments or goals change. We refer to this ever-shifting scenario [...]
Improving Robotic Exploration with Self-Supervision and Diverse Data
Abstract: Reinforcement learning (RL) holds great promise for improving robotics, as it allows systems to move beyond passive learning and interact with the world while learning from these interactions. A key aspect of this interaction is exploration: which actions should an RL agent take to best learn about the world? Prior work on exploration is typically [...]
Combining Offline Reinforcement Learning with Stochastic Multi-Agent Planning for Autonomous Driving
Abstract: Fully autonomous vehicles have the potential to greatly reduce vehicular accidents and revolutionize how people travel and how we transport goods. Many of the major challenges for autonomous driving systems emerge from the numerous traffic situations that require complex interactions with other agents. For the foreseeable future, autonomous vehicles will have to share the [...]
Causal Robot Learning for Manipulation
Abstract: Two decades into the third age of AI, the rise of deep learning has yielded two seemingly disparate realities. In one, massive accomplishments have been achieved in deep reinforcement learning, protein folding, and large language models. Yet, in the other, the promises of deep learning to empower robots that operate robustly in real-world environments [...]