The object was to design a reaction wheel system capable of generating torque to robustly balance the system.
Sponsored By: PSU Applied Research Lab
Team Members
Donovan Gorgas | Tyler Bos | Haram Kweon | Gregory Schweiker | Geran Triano | | | | | | |
Project Poster
Click on any image to enlarge.
Watch the Project Video
Download the Project Summary
Project Summary
Overview
The goal of the project was to successfully balance an unstable two-axis inverted pendulum using induced torque from two mounted flywheels. These flywheels are powered by two motors selected by the sponsor that have the specifications to produce the required torque. A 9-DOF (degree of freedom) inertial measuring unit (IMU) is on top of the pendulum to track movement in each axis, and this data is sent to an Arduino that contains a reinforcement learning model. This reinforcement learning model will then provide the necessary outputs to successfully keep the pendulum balanced.
Objectives
-To design a reaction wheel system capable of generating torque to robustly balance the system
-Integrate new high torque motors into the current prototype and refine Arduino code
-Finalize machine learning algorithm to provide correct stabilizing inputs to the Arduino
Approach
-Conducted basic research on other two-axis balancing systems by graduate teams
-Considered three different flywheel geometry before settling on a wheel with spokes
-Created MATLAB scripts to find flywheel moment of inertia and necessary system balancing torque
-Obtained system constraints to finalize flywheel geometry with sponsor’s approval
-Refined SolidWorks modelling for system with new motors, motor holders, and flywheels
-Project was continued from last semester; all the requirements were given and last year’s work to us (ex. Sample AI code guideline and new motor (Maxon EC 60 flat) motor and new motor controller (ESCON 70/10))
-Conducted test work with last year’s motor and motor controller with IMU input given by Arduino code to produce stabilizing motor torque
-OpenAI gym with TensorFlow was used to simulate the system dynamics.
-Created PID (proportional-integral-derivative) controller with python coding
-Tested the new motor controller with MAXON ESCON studio and encountered hall sensor error
Outcomes
-Future teams will have detailed documentation that will allow them to continue the project
-The sponsor will have a completed SolidWorks model and a quote for costs involved with machining the flywheels
-A completed PID controller