Here are a few of the projects I have worked on over the years. If you would like to see additional videos look at my youtube channel.

Comprehensive Automation of Specialty Crops

I worked with a team from Carnegie Mellon University on a USDA initiative called the Comprehensive Automation of Specialty Crops. This is a large multifaceted project involving multiple universities and industry partners. I worked with a smaller subset of the project focused on developing an autonomous vehicle for use in orchards. There are a variety of uses for an autonomous vehicle in an orchard such as data gathering and precision spraying. I worked specifically on developing the navigation and control of the vehicle. My particular focus was improving the ability of the vehicle to leave an orchard row, turn around and enter the next row. Due to the tight spacing of rows in a modern orchard a single constant curvature arc would not allow the vehicle to make it into the next row. To solve this issue I had to develop a method of generating 3-point (sometimes referred to as K) turns. In addition to this requirement in order to reduce stress on the vehicle the turns had to be smooth (of constant curvature). The paths also had to be able to be generated quickly due to the vehicles constantly changing knowledge of the environment. To create feasible paths for the vehicle to follow I created a model predictive optimal planning method. I generated parameterized arcs that I optimized using a cost function and a general non-linear optimization method. Using this method I could generate feasible paths between any two arbitrary vehicle locations. To generate a 3-point turn three of these paths would be combined. I then leveraged the design of the paths to create a simple control strategy to keep the vehicle on the paths. This method was tested in local orchards around Carnegie Mellon in preparation for a two week field test in Washington State. During the field test the vehicle successfully drove over 20 km autonomously with no user interaction or failures. Below are a series of videos from those tests. In addition I wrote my master’s thesis on this work and it can be found here.

In this video the autonomous vehicle can be seen demonstrating a 3-point turn.

This video is a half hour demonstration of the vehicle navigating an orchard. It is sped up 10x. The vehicle can be seen making a successuful 3-point turn at the end of each row. Additional Videos can be seen on my youtube channel.

Actuator for a Throwable Jumping Robot

During my senior year at Johns Hopkins University I participated on a group industry sponsored project. The project my team worked on was proposed by the JHU Applied Physics Lab. The sponsors wanted our four student team to design an actuator for a throwable jumping robot. The robot that they intend to use the actuator in is the size of a softball but shaped like a soccer ball. Each black panel on this soccer ball shape would be a panel that pops out propelling the robot. A linear actuator would power each of the twelve black panels. To create such a robot each linear actuator needed to fit inside a cone one and a half inches high and have a base diameter of an inch. This seemingly impossible task proved even more difficult than at first glance. Pneumatics, solenoids, and linear motors - the traditional means of producing a linear force - proved impossible to use. Instead we had to design a unique three stage system using pager motors, a compression spring, and an extension spring that fires, retracts, and then reloads internally. No actuator this small had ever been designed to output as much force as this actuator. As part of the team I developed a CAD design for our proposed actuator. We then had rapid prototype parts made. Our design called for a miniature gearbox. In order to create the necessary gears we had to use a wire edm machining process. We designed and ordered a custom spring to meet our exact energy storage and stiffness specifications. Once all the components had been assembled we built a prototype and began testing it. Iteratively testing and modifying the design we showed that our design could work however more reliable and preferably stronger motors would be required. At the end of the school year we presented our work in front of a panel of professional engineers who were impressed with our work. This was a very exciting project to work on partly because it was challenging, but also because it gave me the chance to work with a great team of students, as well as receive criticism from experienced engineers. Below are three videos. The first is a simplified version showing a conceptual version of our mechanism. Below that is a full 3d cad rendering of the complete design. Lastly is a clip of the prototype jumping.Also a complete copy of our final report can be found here as well as a short press clipping here.

A demonstration of our three stage linear actuator. First it pulls the ring down compressing the main spring. In the second stage the base is released causing the main spring to decompress. For the final stage a smaller weaker spring is used to pull the entire mechanism back inside the robot.

In this video the complete model can be seen.In the rear of the model are the motors and gear box. In the center key items that can be seen are the vertical beams and clips which control the 3-stages of the actuation

In this video a demo of a working prototype can be seen. It is using a weaker spring than specified so it doesn't jump very high.

This is a team video we made for a schoolwide competition. It details our work up until the point it was made (just before prototype testing).

Touch Thimble

Through an REU grant I was given the opportunity to work in Allison Okamura's Haptics Lab at Johns Hopkins on a research project. Dr. Katherine Kuchenbecker, who was the primary investigator on the project, was interested in the importance of contact information in haptic (touch) feedback. Contact information refers to a person's knowledge of their contact with something. This is in contrast to force information, which refers to a person's knowledge of how much force is being applied to them. To investigate contact information I developed a new end-effector for a haptic device. A haptic device is essentially a robotic arm connected to a computer. The computer tracks the movement of the arm and translates that motion into motion of a sphere or other virtual representation in a virtual environment. The computer then calculates forces based upon interactions of the virtual repesentation and the environment and outputs that force to the user. A traditional interface for haptic devices, is a thimble that grips a person's finger tightly. The thimble, because it grips a finger tightly, only presents force information to the user. The "touch thimble" that I developed added contact information to the thimble concept. We conducted a user study using both the normal and touch thimble to determine the role of contact information, but our results proved inconclusive. Details regarding the touch thimble, experimential procedure, and results can be found in the paper here.

This is a picture of the touch thimble attached to a haptic device.

This is a CAD drawing of the touch thimble. The red part is a ring that goes around a person's finger, the white cubes are foam blocks that act as springs and the blue part is plastic shaped like a person's finger and contacts a person's finger tip when a force is applied.