Current Projects
Object Tracking and Image Separation Based on Color, Pattern, and Shape
BarnTech [Patent Pending EFS ID: 7864431] is an embedded system that adheres to the philosophy that the best user interface to any system is neither seen nor heard. The objective of the system is to capture the motions and events of an equestrian rider for training pruposes.
Hardware components include an HD and CCD camera, high-torque ServoCity servos, Sony LANC interface via ELM624 (providing zoom, focus, and standby capabilities), ATMEGA328, and a PC-based kiosk with touchscreen.
A Haar classifier is trained to detect a horse and rider. OpenCV runs the Haar classifier algorithm along with Lucas-Kanade and predictive optical flow algorithms to determine a riders relative x,y coordinates. A communication protocol exists between OpenCV’s analysis and the RX end of the ATMEGA328. Angle adjustments are calculated and sent to the servo. The result is an automated recording of a rider constantly optimized to the center of screen.
BarnTech-Diagram-v1.ppt
Optimizing Joint Placement for Bipedal Locomotion and Control
This independent research project ("The Peter Rush Project") exists to further enhance my understanding of bipedal locomotion. I am focused specifically on designing mechanics, interfaces, and algorithms to overcome the inherent challenges found in biped development; dynamic load balancing, actuator network communication, volatile terrain navigation, and power consumption. The skeletal system is constructed from aluminum 6061 and houses a series of high-torque/high-speed linear actuators. Each actuator has a custom low power dissipation motor controller network node that independently communicates with neighboring sensors and reports geometric positioning data back to a central control unit. The current weight of the system is ~36kg at a height of ~1.5m.
Using Air Muscles to Build Biologically Inspired Robots
Mimicking the biological muscle structures found in humans and canines with air pressure, silicone, and braided nylon is an efficient way to generate realistic human-like behavior. It has proven to be a robust means of locomotion ideal for applications that demand such impulsive behavior, e.g., jumping, quick muscle adjusting. Variable amounts of air pressure are controlled through a manifold of air values and solenoids. In most scenarios a minimum of two muscles are used per joint (one interior and one exterior to the angle).
----
|