ECE 4902 Spring 2018 Group 1829 - Unmanned Air Vehicles Final - - PowerPoint PPT Presentation

ECE 4902 Spring 2018 Group 1829 - Unmanned Air Vehicles Final Presentation

Tony Chen (EE) David Kay (EE) Ravi Patel (CSE/EE) Sponsor: UConn ECE Dept. (Prof. Shalabh Gupta)

Agenda

• Our Task
• Component Selection
• Imaging
• Progress

○ Hardware ○ Software

• Timeline
• Costs

Our Task

To create an drone that can:

• Identify boxes on ground
• Orient over, lower down to box
• Pick up and take off with box

Our Task (cont.)

Basic Components: Frame

DJI Flamewheel F450

• Strong
• Modular

○ Easy to replace broken parts ○ Easy to mount components

• Affordable

Flight Controller

Pixhawk Autopilot

• Open source autopilot
• All-in-one unit

Sensors:

• 3-axis 16-bit gyroscope
• ST Micro LSM303D 3-axis 14-bit

accelerometer / magnetometer

• MEAS MS5611 barometer

Interfaces:

• UART
• CAN
• I2C
• MicroSD

Drift Minimizer

PX4Flow Sensor:

• Optical flow smart camera
• Uses ground texture and visible features to

determine aircraft ground velocity ○ Works where GPS can’t reach

• Indoors and outdoor, including low light

conditions (no need for external illumination)

• 3 axis gyro allows for compensation of vehicle

tilt and heading relative to the ground

• Designed to work with Pixhawk FC

Target Identification

High Contrast Color Combinations:

• Easily Detected
• Limited Interference

Alternatives: Downsides: QR Codes Requires Very Clear Image Retroreflective Material Expensive / Single Color Shapes Requires Very Stable Image Text Computation Heavy

Target Detection: Pixy Cam

Pros

• Open source
• One-touch object and color learning
• Color for object detection
• Up to seven color signatures.

○ Red, Orange, Yellow, Green, Cyan, Blue, Violet

• Multi-color combinations
• Detects angle, width, and height

Cons

• White balance
• Incandescent vs Fluorescent lighting can affect color

signature ○ Can be easily recalibrated for each lighting

Onboard Computer

Alternative: Raspberry Pi3

• 1.2 GHz Quad-core CPU
• 1GB RAM
• Micro SD slot
• Raspbian (Debian branch-off)
• $35

ODROID XU4

• 2GHz octa-core CPU
• 2GB LPDDR2 RAM
• Micro SD slots
• Linux 16.04 or Android
• Better for onboard flight computations
• $61

Odroid XU4 Raspberry Pi3

Signal Flow Diagram

Pixy CMUcam Sensor

Flow Sensor

Electronic Speed Controller Gripper

Hardware Design - Gripper

Gripper:

• Two Servo Design
• 3D Printed
• Traction Tape

Landing Gear:

• Separate Component
• Drone Safety

Hardware Design

Complete:

• Flight controller wired

○ Stable flight

• Flow sensor installed

○ Minimized drift

• Gripper Installed

○ Bluetooth connectivity In-Progress:

• Tuning

○ Increase smoothness & flight behavior ○ Improve altitude hold (sonar)

• Pixy/CMUcam5 control Integration

○ Adjust position based on CMUcam position data

Software Design

Sensor Data

XU4 Gripper CMUCam5 Sensor

Switch Data Commands

Software - Image Processing

Arduino Serial Output PixyMon Color Signatures

Test Flight - Image Processing

Test Flight - Box Manipulation

Next Steps: Optimizations

• Two Pixy Logic --->Accurate orientation over object
• Arduino port to Odroid ---> Faster communication
• Kinect IR/Depth/Flat ---> Better altitude accuracy
• Conversions/Flight Controller ---> Data translation

Timeline

April 21-27: Demo-Day Preparation March 25-31: Complete Software (Controls Test) April 8-14: Drone Optimization and Aesthetics April 1-7: Complete Software (Communication Test) April 15-21: Presentation Finalization (Report / Poster) Record Demo Videos

Costs To-Date

Items Cost

Pixhawk Flight controller

$110.00

PX4 Flow Sensor

$150.00

Odroid XU4

$61.00

CMUcam5 (Pixy)

$69.00

Kinect Sensor

$50.00

Gripper

$20.00

Receiver

$34.00

Power Module

$22.00

Questions?