Timeframe: June 2019 - December 2019
Part of Modlab (member of GRASP LAB)
The Modboat is a modular robotic boat powered by a single motor. The motor spins a bottom plate to which are attached two free-swinging flaps. By precisely controlling the rotation of the bottom plate, it is possible to make the robot swim forward and make turns, even in place. This represents the first step towards modular floating sysetms by creating a low-cost, simple, and maneuverable modular unit.
The Modboat is controlled by an ESP32 attached to a lasercut frame
contained in an acrylic tube. The flipper asembly on the bottom is
3D printed. A central computer runs a MATLAB script which receives
position data from a motion capture system and then transmits the
parameters for the swimming motion necessary to reach the desired
waypoint.
During the project, I was part of a two person team, designing and
fabricating the flippers and parts of the frame, programming the
swimming behavior, running experiments, and analyzing telemetry data
to find a piecewise linear relation between one of the swimming
parameters and robot turning rate.
Timeframe: June 2019 - December 2019
Part of Modlab (member of GRASP LAB)
Quori is an open-source, low-cost robotic platform for allowing researchers to study Human-Robot Interaction. It has a holonomic base, an actuated waist, a pair of 2-DOF arms, and a rear-projected face. Sensing is acheived through a mix of RGBD cameras and LIDAR. In this specific implementation, Quori is capable of turing to follow the closest person, mimicking the arm motions of a person, bowing, dancing, and emoting with different facial expressions.
As part of an exhibit at the Philadelphia Museum of Art about
the future, the Modlab contributed Quori, to interact with patrons
at the museum. The exhibit is Designs for Different Futures, on
display from October 22, 2019 to March 8, 2020. Click
here
for an article about Quori in the exhibit.
Over the
course of five months, I worked as part of team to add perception,
decision making, and physical behaviors to Quori. My roles included
programming the physical behaviors using trajeory generators and
a robotic puppet, managing the git repository, creating
documentation for the various software and hardware components,
tracking team members' progress, setting team goals, and planning
meetings.
Timeframe: August 2018 - June 2019
Part of ScalAR Lab (member of GRASP LAB)
In order to assess the affects of delay on swarm robotic systems,
I generalized an existing swarm model by adding collision avoidance
to it and characterizing the new swarm behavior through simulations
I performed in MATLAB.
I also conducted mixed reality experiments whereby a few
physical robots interacted with many virtual agents.
In these experiments, the
robots were miniature Autonomous Surface Vehicles (mASVs), small
differential drive boats, that moved about in a water tank.
I tuned a
PID controller and used a feedback linearization to allow the boats
to respond to holonomic motion commands. I used a motion capture
system to provide localization and tracking. The tracked positions
of the boats was fed into the simulation which would in turn give
commands to the boats after calculating the interactions between
all agents, physical and virtual. The physical and virtual
agents moving in sync validates the findings from previously
conducted simulations.
In order to visualize the swarm interactions, I wrote a program
using MATLAB's computer vision toolbox to combine video footage of
the experiments in the water tank and video output from the
simulation as seen in the picture. During this process, I calculated
camera intrinsic parameters and then created a system for deriving
the camera extrinsic parameters for each video which could be filmed
from a number of positions.
Timeframe: September 2015 - June 2016
Part of Bioinspired Robotics and Design Lab
To facilitate the development of the soft robotic actuator listed in the next section, I designed and built a mechanical tester to speed up data collection of the properties of the various actuator designs we were investigating. Consisting of a motor actuated linear stage and a force sensor, the mechanical tester was designed to be simple, cheap, and easy to build so that it could be used by hobbyists wishing to get involved with soft-robotics.
I wrote an paper for the Robot Makers 2 workshop at Robotics: Science and Systems 2016 where I presented.
View the paper
Timeframe: June 2015 - June 2016
Part of Bioinspired Robotics and Design Lab
Working with a PhD student, we set out to make a soft robotic actuator
to act as a leg in a quadrupedal robot. To this end the actuator had to
be easy to manufacture, have a high strength to weight ratio, possess
two degrees of freedom, and be capable of large deflections.
The actuator we settled on took the form of four silicone tubes attached
in the shape of a three-pointed star and could be made with a single
casting.
Using three of these actuators, we made a dexterous gripper capable of
rotating objects in its grasp with out the need of a wrist as well as
handling a variety of objects, from heavy to fragile, with the same
control scheme.