Maryland Robotics Center, University of Maryland, College Park, Maryland, USA

[Airicist]

Website - robotics.umd.edu (https://robotics.umd.edu)

youtube.com/UMDRobotics (https://www.youtube.com/UMDRobotics)

linkedin.com/company/marylandroboticscenter (https://www.linkedin.com/company/marylandroboticscenter)

instagram.com/marylandroboticscenter (https://www.instagram.com/marylandroboticscenter)

Director - Derek Paley (https://www.linkedin.com/in/derek-paley)

Projects:

Baxter robots, learning to think, see, and cook (https://pr.ai/showthread.php?13901)

robotic cleaning (https://pr.ai/showthread.php?13911)

Robo Raven (https://pr.ai/showthread.php?1600)

RoboTerp (https://pr.ai/showthread.php?6825)

Optical Hand (https://pr.ai/showthread.php?6826)

RoboCrab (https://pr.ai/showthread.php?5578), horseshoe crab

[Airicist]

https://youtu.be/mjOWpwbnmTw

Published on Apr 29, 2013


Complete individual wing control allows crazy aerobatics like back flips and dives, previously not possible with mechanical birds! Hawk attack at 1:49.

[Airicist]

https://youtu.be/BpB-dmIP8ig

Model-Predictive Asset Defense by Team of Autonomous Surface Vehicles

Published on Oct 29, 2013


In this video, we present a contract-based, decentralized planning approach for guarding a valuable asset by a team of autonomous unmanned surface vehicles (USV) against hostile boats in an environment with civilian traffic. The particular objective for the team of USVs is to maximize the expected time it takes a hostile boat to reach the asset. The team has to cooperatively deal with uncertainty about which boats poses an actual threat, employ active blocking to slow down the movement of boats towards the asset, and intelligently distribute themselves around the asset to optimize their future guarding opportunities. The developed planner incorporates a contract-based algorithm for allocating tasks to individual USVs through forward simulating the mission and assigning estimated utilities to candidate task allocation plans. The task allocation is based on marginal cost based contracting that allows decentralized, cooperative task negotiation among neighboring agents. The planner is capable of computing task allocation plans in real-time and is general enough to be used for a variety of scenarios. The underlying behaviors that correspond to individual tasks are optimized for two specific mission scenarios.

E. Raboin, P. ?vec, D. Nau, and S. K. Gupta. Model-Predictive Target Defense by Team of Unmanned Surface Vehicles Operating in Uncertain Environments. IEEE International Conference on Robotics and Automation (ICRA '13), Karlsruhe, Germany, May 6-10, 2013.

[Airicist]

https://youtu.be/ydnjS__8Ooc

Published on Mar 16, 2014


This video presents an imitation learning approach for a fluid pouring task, which consists of grasping a bottle containing a fluid and pouring a specified amount of the fluid into a container placed on a rotating table. The robot learns how to do this task from human demonstrations. In addition to learning from successful demonstrations, our approach allows learning from errors made by humans and how they recovered from these errors in subsequent trials. We collect human demonstrations of the task and label each
demonstration as either a success or a failure. A tracking system
is used to automatically extract motion parameters from video
recordings of each demonstration. We present an algorithm
that combines support vector machines (SVM) based classifiers
and iterative search to generate initial task parameters for the robot. If the robot fails to perform using these parameters, we learn simple rules to refine them by capturing how human demonstrators change parameters to transition from failed demonstrations to successful demonstrations. We report experimental results consisting of a 5 DOF robot using the learned parameters to successfully perform the pouring task to illustrate our approach.

[Airicist]

https://www.youtube.com/watch?v=L-J3NW3sXY8

RoboCrab: A Horseshoe Crab Inspired Amphibious Robot for Righting in Surf Zones

Published on Mar 23, 2014


This video presents RoboCrab, an amphibious robot capable of traversing moderate surf zone environments. By taking inspiration from the morphology, locomotion, and righting behaviors of a horseshoe crab, the robot is designed for traversal and righting on granular terrain, open water, and turbulent surf zones.

[Airicist]

https://youtu.be/TEMfe7_NKiY

R2G2: Camouflaging, pipe-inspection, and high traction capabilities

Published on Apr 25, 2014


This video demonstrates R2G2 camouflaging itself in the woods, performing pipe-inspection tasks, and slithering over tile floors and rocky terrain.

[Airicist]

https://youtu.be/3d-qSEAckmc

Bio-inspired Robotics (ENME 489L) Fall 2013

Published on Apr 25, 2014


This video shows the robots built by the students as a part of the bio-inspired robotics (ENME 489L) course taught during Fall 2013.

[Airicist]

https://youtu.be/qg0kjzrag1I

Using Failure-to-Success Transitions in an Imitation Learning Framework

Published on Jul 2, 2014


The video demonstrates an imitation learning approach for a dynamic fluid pouring task. Our approach allows learning
from errors made by humans and how they recovered from
these errors subsequently. We collect both successful and failed
human demonstrations of the task. Our algorithm combines a
support vector machine based classifier and iterative search
to generate initial task parameters for the robot. Next, a
refinement algorithm, capturing how demonstrators change
parameters to transition from failure to success, enables the
robot to address failures. Experimental results with a Baxter robot illustrate our approach.