Autonomous Waste
Sorter
Group 9
Yankarlos Bencosme
Matthew Madden
Kyle Beuttel
Michael Bosse
Meet The Team
Yankarlos Bencosme - Computer Engineering major interested in embedded and
real time system programming.
Kyle Beuttel - Electrical Engineering major interested in system engineering and
Test and Integration engineering
Michael Bosse - Computer Engineering major interested in reverse engineering
and cybersecurity.
Matthew Madden - Electrical Engineering major interested in research and
development.
Problem Statement
❏ Roughly 75% of America's waste is recyclable; however, they estimate that only about
30% of that waste is actually recycled.
❏ The U.S. recycling rate is around 32%. If we can devise a method to increase that rate
to 75%, experts say the effect will be equivalent to removing 50 million internal
combustion vehicles from the roadways.
❏ Given its importance, why is our recycling percentage low?
Project Description
❏ Create a system that is integratable into multiple types of waste baskets and
cans.
❏ Instead of having separate baskets for waste and recyclables, we will now
have one.
❏ Electronic system that sorts recyclable materials from waste.
❏ Sorting will be done using trained AI models and computer vision technology.
Problems to Consider
❏ Waste baskets are typically outside or in places where power outlets are generally not
available. For our product to be successful and at convenient as possible, how can we
combat this?
❏ In high traffic areas, our systems needs to be able to work quickly to handle constant trash
inputs.
❏ When people have multiple items to dispose of, can the device handle multiple items?
❏ Since this is a waste basket, will the device suffer of failure from waste contamination such
as a dirty camera lens?
Goals
❏ Main Goal - Create a system that is able to identify and sort at least 3 different common
recyclables.
❏ Advanced Goal - Power the system using a portable power source such as batteries.
❏ Advanced Goal - Implement a buffer system to allow multiple items to be inserted.
❏ Stretch Goal - Implement a solar powered system that utilizes a rechargeable battery pack.
❏ Stretch Goal - Have the system complete the identification and sorting process in 5 seconds or
under.
Specifications
Specification Target
Time to successfully complete the acceptance and <= 15 Seconds
sorting process.
Accuracy of sorting process and overall success >= 80%
rate
Number of Accomodated Materials 3 Different Materials
Storage capacity >= (3) 5 gallon Baskets
Final product dimensions <= 25”W x 46”H
53 Gallon Capacity
Hardware Diagram
Prototype Design Concept
Prototype Design Concept (Top View)
Extended (Hold Retracted (Release
Position) Position)
Prototype Design Concept
Servo Motors
DS3245SG
SF3218MG Cost $15
Cost $2 Operating Voltage 5V (4.8-7.2V tolerant)
SG90
Operating Voltage 5V (3-6V tolerant) Torque 20.5 kg per cm
Cost $50
Torque 1 kg per cm Operating Speed 0.18 sec/60°
Operating Voltage 5V (5-8.4V tolerant)
Operating Speed 0.12 sec/60° Degrees of 270°
Torque 35 kg per cm
Rotation
Degrees of 180°
Operating Speed 0.18 sec/60°
Rotation
Degrees of 180°
Rotation
Linear Actuators
Design
Selecti
on
Cost $189
Cost $50 Input Voltage DC 12V
Cost $25 Input Voltage DC 12V Speed 0.19in / sec
Input Voltage DC 12V Speed 0.55in / sec Protectional IP54
Class
Speed 0.6in / sec Protectional IP54
Class Actuation Force 6000N
Protectional IP54
Class Actuation Force 1000N Max transfer Rate 400 FPS
Actuation Force 20N Max Load 225 lbs Max Amp. Draw 9A
Max Load 4.5 lbs Max Amp. Draw 3A
Max Amp. Draw 0.1A
H-bridge Controller
DRV8833 DRV8876 TB6612FNG
Voltage 2.7 - 10.8v 4.5 - 37v 2.5 - 13.5v
Amperage 2A per channel 3.5A per channel 1A per channel
H - Bridge Dual H-Bridge Dual H-Bridge Dual H-Bridge
Low Power(LP) LP Mode Ultra LP LP mode
Cost $1.79 $1.94 $1.97
Linear Actuator
❏ Requires 12V to extend at full speed (lower voltage can be used but at slower
speeds)
❏ Polarities need to be reversed in order to retract the actuator.
❏ Polarities are effectively reversed by using an H-Bridge
Camera
MT9V03
OV564 Cost 4 $65
Cost
0 $109 Array Size 752 x 480
OV772
Cost
5 $20
Array Size 1600 x 1200 Power 140mW @ 60FPS
Requirement
Power 160mW @ 60FPS
Array Size 640 x 480 Output Format Raw RGB Data
Requirement
Power 120mW @ 60FPS Output Format YUV
Requirement RGB
Lens Size ⅓”
Raw RGB Data
Output Format Raw RGB Data Max transfer Rate 400 FPS
GRB 4:2:2 Lens Size ¼”
RGB565/555/444
Max transfer Rate 60 FPS
Lens Size ¼”
Max transfer rate 60 FPS
Peripheral Board
Raspberry Pi 3B+ OpenMV H7 OpenMV H7 Plus
Raspbian OS No OS No OS
Software OpenMV or OpenCV OpenMV OpenMV
Storage Memory ===== 1GB 2MB Flash 32MB Flash
Camera External Camera (640x480) 75fps (2592x1944) 50fps
Design Closed architecture Open Architecture Open architecture
Cost $60 $85 $90
MCU
❏ ATMEGA328P-PU is low cost with a sufficient number of GPIO pins and PWM channels.
❏ Programmable using Arduino IDE and an in-system programming device (pins integrated into PCB).
❏ Supply voltage supports 5V which is the same range for servo. No need for an additional step-down
IC.
Supply PWM
Name Clock Flash GPIO Cost
Voltage Channels
ATMEGA328P-
20 MHz 32KB 1.8V-5.5V 23 10 ~$3
PU
STM32F100C4 24 MHz 16 KB 2V-3.6V 37 9 ~$5
MSP430G2553 16 MHz 16 KB 1.8V-3.6V 24 8 ~$3
Design Implementation
Voltage Regulator
❏ Linear actuator requires 12V to operate fast enough to reach timing specification.
❏ MCU and servo are rated for 5V supply.
❏ Using a linear voltage regulator, we step-down 12V RAW supply to 5V for MCU and
servo.
PCB Schematic
PCB Board
Power Supply - Battery or Wall Outlet
❏ Voltage required to power system is 12V with an estimated peak current draw of
2.5A.
❏ Opted for a 12V barrel jack power supply.
❏ Allows for use of 12V DC-DC 3000mAh rechargeable battery.
❏ Can also use AC-DC 12V 3A power supply as a secondary option.
Input Power - Barrel Jack & Micro USB
❏ Using a barrel jack rated for 3A because estimated peak current draw of system is
2.5A
❏ Output power headers integrated into PCB design for a micro USB connection to
power peripheral board. Single power source for system.
Barrel Micro
Jack USB
Software Diagram
Software
❏ We will be using the OpenMV IDE since our design is based on their H7 Plus camera.
❏ The programming language used is python.
❏ Features a viewer that allows us to see what the camera is seeing and if its
recognizing the image.
Software
❏ We will be using Edge Impulse to train our object detection AI model.
❏ Integrates seamlessly into the OpenMV IDE which allows training data to be
uploaded effortlessly.
❏ Provides multiple levels of training to accommodate different chip performance
specifications.
Software
Software
Work Distributions
Work Distributions (cont.)
Budget Estimations
Description Quantity Estimated Estimated
Cost Total
PCB 2 $120 $240
Camera 1 $60 $60
Servo 1 $50 $50
Actuator 1 $30 $30
Housing and Release Mechanism 1 $40 $40
Estimated Total - - $420
Constraints
❏ Cost - We have decided to stay under $500 total to design, manufacture, and test
our system.
❏ Time - We only have roughly 25 weeks to research, design, and develop the system.
❏ Space & Size - We want our system to be able to integrate seamlessly into existing
common waste baskets that are typically seen in public.