About Us

The WPI Humanoid Robotics Laboratory houses the 6-foot, 400-pound Atlas Robot from Boston Dynamics. Our robot, named WARNER for WPI Atlas Robot for Non-conventional Emergency Response, found its way to WPI for the DARPA Robotics Challenge. WPI was the 1st university to take delivery of an Atlas robot in 2013 and it has been here ever since. Team WPI-CMU came in 7th in the 2015 DARPA Robotics Challenge Finals.


Michael Gennert
Director, Robotics Engineering
Professor, Computer Science
Professor, Electrical and Computer Engineering

Jie Fu
Assistant Professor, Robotics Engineering
Assistant Professor, Electrical and Computer Engineering


NASA Space Robotics Challenge

Team WPI Humanoid Robotics Lab has successfully cleared the qualifiers for NASA Space Robotics Challenge. In this project, the team is programming Valkyrie to perform repairs of a habitat on Mars caused because of a sand storm in a simulated environment. More details about the competition are available here. To know more about the team and current development of this project... Read More
Shadow Motion For Atlas
Initially, we are trying to develop an algorithm which will make the ATLAS mimic human body motion using Microsoft Kinect Sensor. The idea is to have a concrete mapping between human joints and ATLAS joints. The human joints are obtained by Skeletal Tracking. Later we will ... Read More
Facial Recognition using Sparse Data Representation on an Open Source Neural Network Model.
This project is an open source face recognition library based on neural networks in java using Walnut AI, an algorithm and training data found on matlab, and using OpenCV as a prefilter to single out areas of interest ... Read More

Robot Learning

The purpose of the project is to mimic human’s learning and motion mechanisms in order to create an adaptive walking gait on Atlas Humanoid Robot. The project tackles the issue of lack of generalization that current learning mechanisms face by remembering past experience, simulating risk averse behavior, and significantly reducing state spaces. In addition, the project builds the entire biologically inspired motor hierarchy which allows for great level of adaptability during motion tasks, including, but not limited to walking.
Footstep Planning
Footstep planners generate a sequence of steps for repeating a predefined walking motion of robot, which is used for navigation of biped humanoids. These planners obviate the need of computationally expensive process of generating full body motions for navigation... Read More
Robots have historically been tasked with work requiring a variety of fine motor skills, such as welding and packaging. While these applications have generally been industrial in nature, robots have more recently been applied in areas of artistic interest. This project aims to examine the use of the Atlas humanoid robot to process an image and produce a painting. Through the use of OpenCV for contour detection, G-code generating algorithms, and manipulation of painting mechanics, our team will demonstrate a calculated approach to what is considered a largely subjective task.

Dual Arm Manipulation Control of ATLAS

The goal of this project is to create a task specific controller over the existing full body controller in ATLAS that provides overall stability to the robot. The controller aims to provide synchronized arm manipulation which is required by tasks such as... Read More
Balancing Inverted Pendulum
Abstract—The purpose of this project is to create a sub-controller that can be merged with the existing full-body controller used by Atlas so as to balance an inverted pendulum by using his hand. The addition of this controller will give Atlas a sense of balancing things... Read More
WPI - Toyota Project
We live in very exciting times. What once was science fiction is now becoming reality: Self-driving cars, cars that are aware of their passengers and dynamic external environments, smart cars that may truly successfully accomplish safety for all. However, still missing is a better understanding of ... Read More

Deep Q-learning For Walking In Humanoid Robots

Existing methods for walking created for humanoid robots suffer from a lack of adaptability to new environments, due to their reliance on using only higher level motion control with relatively fixed sub-motions, such as taking an individual step. They also require a large amount of ... Read More


Sat, Jun 11

TouchTomorrow ATLAS Exhibit

Wed, May 4

WHRL Cookout!


Coming Soon


WPI Humanoid Robotics Lab
100 Institute Road
Worcester, MA - 01609

email us at whrl-admin@wpi.edu