cse 424 final presentation team members: edward andert shang wang michael vetrano thomas barry roger...
TRANSCRIPT
Team Seek And Destroy
CSE 424 Final Presentation
Team Members: Edward Andert
Shang Wang
Michael Vetrano
Thomas Barry
Roger Dolan
Eric Barber
Sponsor: Aviral Shrivastava
Overview Scope Schedule Tower Technology and Design Car Technology and Design Video Tower Results and Lessons Car Results and Lessons Conclusion
Outline
Two completely autonomous systems working in parallel
Turret aims at target and decides when to fire
Car follows a person and implements obstacle avoidance
Multiple uses and implementations◦ Military applications◦ Autonomous systems and vehicles◦ Image recognition◦ Children's toys
Overview / Motivation
Car and turret remain autonomously controlled Car detects objects in its path Car responds to overrides from controller and
voice Turret fires at circular objects with easily
defined center point Turret is more responsive and accurate with
pan and tilt system Easily demonstrated with crowd appeal
◦ Very interactive◦ Marketable
Scope
Schedule vs. Execution - Semester 2
Jan Feb Mar Apr
Week 3 4 1 2 3 4 1 2 3 4 1 2 3 4
Stage 2 Optimizations
Interface with sensors
Implement software
Code Documentation
Demonstrations
Developer’s Guide
Website Building
Red - Car Team; Green - Tower Team; Blue - Both Teams; Yellow - Difference (Additional Time Spent); Aqua – Difference (Late Start);
Kinect◦ Camera information used to calculate current position of target◦ Depth information used to calculate time to target
Arduino-USB interface to servos◦ Changes position of turret
Dream Cheeky Turret◦ Fire command sent through USB
OpenCV and OpenKinect libraries used for image processing◦ Filter image color◦ Noise Reduction◦ Detect edges◦ Find Contour
Technology - Tower
Optimize existing tracking algorithm◦ Enhance robustness◦ Add motion prediction◦ Improve performance
Create a more stable tower◦ Kinect needs less interference from moving
servos◦ Revamp old turret design for better stability
Design - Tower
Re-use hardware from previous team◦ Traxxas RC car◦ Atom Board◦ Microsoft Kinect◦ Arduino Mega (Custom DIYdrones model)
Use Microsoft software for ease of interfacing with hardware◦ Windows 7 OS◦ Visual Studio for development (C#)◦ Microsoft Kinect with Kinect SDK
Technology -Car
Start from scratch on software Progressively add autonomous functionality to
the car Create a fail safe to override autonomous
controls Make car follow humans
◦ Build in human recognition using Kinect SDK◦ Demonstrate autonomous tracking with human
interaction Use voice commands to change car behavior
from user◦ Eliminate need for any controller/remote
Design - Car
Recognizing and tracking objects◦ Able to recognize an object in a certain color
Using shape analysis mechanism to filter the background noise and stabilize the turret
◦ Able to shoot the objectaccurately and quickly Enabling auto-shoot and manual–shooting Physics calculation Predict the car movement
Result Overview - Tower
Linux development Computer vision technique and OpenCV
programming Version control Module testing Iteration in software development
Lessons Learned - Tower
Implemented autonomous car ◦ Recognizes the skeletal structure of a person◦ Locks onto that person while in view◦ Follows the person
Feedback loop Depth plane recognition for object detection Voice recognition of commands (stop, start)
Results Overview - Car
Autonomous solutions can be unpredictable at times
Fail safe was necessary Feedback loop important for monitoring
car functions Initialization detrimental
to startup Inconsistent servos
require calibration andalgorithmic correction
Lessons Learned -Car
We have created an autonomous car and turret pair
The turret tracks and predicts object motion and shoots at the target
The car recognizes and follows a human with the ability to take voice commands and detect objects
The team overcame many obstacles and learned valuable lessons in cyber-physical systems and project management
Conclusion
Questions?