Projects
This page will soon be updated. More videos would be uploaded.
Failure is an option: Task and Motion Planning with Failing Executions
We propose a framework to address a task and motion planning setting where actions can fail during execution. To achieve a task goal actions need to be computed and executed despite failures. The robot has to infer which actions are robust and for each new problem effectively choose a solution that reduces expected execution failures. The key idea is to continually recover and refine the underlying beliefs associated with actions across multiple different problems in the domain.
Link to the paper Link to the project website Media Coverage
A General Task and Motion Planning Framework For Multiple Manipulators
We present a general TMP framework designed for multiple robotic manipulators. This is based on two contributions. First, we propose an optimal task planner designed to support simultaneous discrete actions. Second, we introduce an intermediate scheduler layer between task planner and motion planner to evaluate alternate robot assignments to these actions. This aggressively explores the search space and typically reduces the number of expensive task planning calls.
Augmenting Control Policies with Motion Planning for Robust and Safe Multi-robot Navigation
This work proposes a novel method of incorporating calls to a motion planner inside a potential field control policy for safe multi-robot navigation with uncertain dynamics. The proposed framework can handle more general scenes than the control policy and has low computational costs. In the proposed approach, we attempt to follow the control policy as much as possible, and use calls to the motion planner to escape local minima. Trajectories returned from the motion planner are followed using a path-following controller guaranteeing robustness.
augmenting_control_policies_experiment_scene.mp4Learning Behavior Trees From Demonstration
Behavior trees are an architecture designed for execution on a robot which makes it possible to represent complex behaviors graphically, enabling non-expert users to program complex behaviors. We built a framework for the robot to learn behavior trees as a new form of policy. The behavior trees are produced by learning from human demonstration.
Chez Homie Project
This project aim to build a fully autonomous pipeline for the robot to complete a delivery task. Customers could order snacks online, and the robot would automatically pick up groceries and navigate through the building to make the delivery.
This project is part of the Magna project, in which more functionalities such as object classification by faster-RCNN are incorporated.
I am responsible for the manipulation pipeline, but I'm also working with my lab mates on building the behavior trees, the faster-RCNN, and some perception problems.
chez_homie.MOV
Object Pose Estimation via Curvature-based ICP for Goal-Oriented Robot Manipulation
Perception is critical for a robot to accomplish localization and manipulation tasks successfully in household environment. Object localization becomes very challenging in such complicated environment. We proposed a curvature-based Iterative Closest Point algorithm, to help the robot localize a target object in the scene. The point cloud data was collected from the Laboratory for Progress's Fetch Robot, and our result shows that we can correctly localize object from a cluttered scene.
This is our course project for Mobile Robotics (EECS 568).