Final Design Review Aziz Deepti Josh Suhail Stuart
- Slides: 28
Final Design Review
Aziz Deepti Josh Suhail Stuart Final Design Review (Week 15) Administration H/w Placement & Casing Imagine RIT Demo Update Trello Plan 04/25 Clean up battery module Prepare for Demo 04/27 Prepare Aruco Markers required for Demo Finish up EDGE 04/28 Final Gate Review 05/01 Poster & Paper Prep. (& Imagine RIT) 04/26 Setup and Batteries 04/27 Poster Submitted (Arm Day & Imagine RIT) 04/25 04/19 IEEE Paper Submitted Screw-in Pi Cam 04/24 Fit housing on hardhat 04/26 Demo Clean-up for Imagine RIT 04/27
Performance vs. Requirements
Current BOM
IMU Recap ● Data from IMU obtained through Teensy ● Using the data to compute angles ● Eliminating the Teensy ● Testing and Calibration
Hard Iron Calibration ● Data distortion due to external strong magnetic field
Test Data Before Hard Iron Calibration
Test Data After Hard Iron Calibration
Vision System: Testing Recap
Translational Data:
Translational Data:
Translational Data:
Rotational Data:
Rotational Data:
Rotational Data:
Person-To-Chair Testing:
Person-To-Chair Testing:
Solving the Stability Problem
The Stability Problem ● Aruco Marker pose data appears to fluctuate between two different solutions ● Fluctuations disrupts the Person-To-Chair math as it is highly dependent on angles ● (Fluctuation Example: The 45º Y Rotation Test)
Suggested Solution ● Compare current Pose Angle reading to past data ● Determine if it is similar or dramatically different ○ If it’s too different from old data, throw it out ○ If it’s similar to old data, average it into the result
The Outlier Detector
The Outlier Detector
Outlier Detection: Example (Not real data)
Outlier Detection: Filtered Angle Data
Outlier Detection: Filtering the Chair Data ● Apply Outlier Detection to angular filter ● Apply Outlier Detection Algorithm on Person-To-Chair data received from Aruco markers - Force aruco markers to “Vote” on the position of the chair ○ (add one point of past chair data to help skew voting towards past results) ● Apply Outlier Detection to smooth resulting chair data
Outlier Detection: Person-To-Chair (3 Arucos)
Outlier Detection: Person-To-Chair (3 Arucos)
Final Demo