et-navswarm 2016 urc poster
TRANSCRIPT
x
x
Extra Terrestrial Navigation with Particle Swarm Optimization Team Members: Joshua Proulx, Nathan Beaulieu, Haley Breedlove, Colin Pellerin, Samuel Wilkinson, Gregory Hatfield,
Brody McClure, Brian Teune, Nolan Dinero, Aaron Rist, Garrett Malagodi, Iskandar Fendi, Ben Yandell, Zachary Plumer Advisors: Professor May-Win Thein, Professor Se Young Yoon, Michael Johnson
Arduino Mega • Runs an LQR control algorithm (programmed in C/C++)
• Analog and digital input/output pins read sensor data
• Sends sensor data to the Raspberry Pi
Raspberry Pi • Runs the PSO (programmed in Java)
• Uses sensor data and the PSO algorithm to select search points and
plans a path to the destination
Xbee • Wirelessly transmits and receives data allowing communication between
bots
• Sends packets of data through a mesh network
• Adds network stability through self healing, self discovery, and dense
network operation
Abstract
ET-NavSwarm’s mission is to design, build,
and test a swarm of autonomous robots.
These will serve as a testing platform for a
graduate research algorithm: Particle
Swarm Optimization (PSO). PSO is derived
from the concept of swarm behavior found
in nature (birds, bees, etc.). By
communicating and using mathematical
principles, the swarm can find the
best/most of a given objective within a
specified area. The swarm operates without
a centralized leader to search for natural
resources, signs of life, or items of interest.
The robots are designed to traverse rugged
terrain, climb 45° inclines, avoid obstacles,
and collect data on position and elevation.
To accomplish this they were made to be
water-resistant and have the longest
possible run time (battery life). The robots’
sensors and electronics were selected to
fulfill the requirements of each algorithm.
As a proof of concept for NASA, ET-
NavSwarm will perform a large scale field
test with these robots, searching for the
highest and lowest elevations.
Design Constraints
Sensors
Inertial Measurement Unit (IMU)
• 10 Degrees of Freedom
• Determines roll, pitch, and yaw
• Determines altitude via barometric
pressure
Global Positioning System (GPS)
• Determines location
Infrared (IR) Sensors
• Detects obstacles up to 5 feet
• Drive up a 45 degree incline
• Traverse rugged terrain
• Water resistant
• 1 hour of run time
• Cost
• Sensors to provide feedback
Particle Swarm Optimization Design
Controls Analysis Software and Autonomy Circuit Analysis
Path Planning Obstacle Avoidance
Description: • Created by Graduate Student Michael Johnson
• Proof of concept for NASA to aid in Space Exploration
• Algorithm created from studying swarms of animals
• Decentralized with no leader
• Each member acts independently towards a goal
How it will be Tested: • Will be tested on Earth using the two functions below:
• The Sphere Function will serve as the first test platform due to
its simplicity
• The Schwefel Function will be later used to test the complexity
Sphere Function Schwefel Function
Desired Path Obstacle
New Path
Description: • Determines the best way around a stationary obstacle
• Obstacles are known prior to the start of the mission
Description: • Avoids stationary obstacles that are in the desired path
• Obstacles are not known prior to the start of the mission
www.unh.edu/navswarm
2015-2016
Original Design: BJT H-Bridge Original BJT H-Bridge Design Problems
• Low Current Capacity (2A)
• Voltage Drop Across Darlington Transistor Pairs
• Slower Switching Times
Updated Design: MOSFET H-Bridge MOSFET H-Bridge Design Problems
• Soldered on Perf-board
• Time consuming to produce
• Prone to wiring/soldering errors
• Difficult to fix errors
Final Design: Pololu 18v17 Pololu 18v17 High Power Motor Driver
• Handles 17 Amps continuous at 12 volts
• Very Small (1”x1”)
• Minimal connections
• Consistent operation
Description: • Method to determine inputs to the motors
• Inputs are based on relative error
• Error is based on position relative to goal set by
particle swarm optimization
• Diverse controller options
• PID and LQR modeled, LQR selected as best
• Based on closed loop control system
Controller Characteristics: • PID (Proportional Integral Derivative)
◦ Simple
◦ Highly susceptible to noise
◦ Loses effectivity at longer distances
• LQR (Linear Quadratic Regulator)
◦ Complicated preparation, simpler application
◦ Inputs calculated before operation occurs
◦ Effective for all ranges
◦ Less controller effort is required
Resulting System: • Travels towards goal based on the
calculated inputs from the control system
• Robots not already facing their goal rotate
until it is more efficient to move forward
• End result is a curved path to location
• New path is faster with higher efficiency
Swarm
Physical
World
Arduino Mega
Controller H-Bridge
(Motors)
State Infrared
IMU GPS
Sensors
Serial
Port
Serial
Port
Raspberry Pi
State
Data & Control Platform
Application
Program
Interface PSO
XBee
IMU Holder IR Sensor Housing
Q1
2N2222A
Q3
2N2222A
U1A
4001BD_5V
U1B
4001BD_5V
U1C
4001BD_5V
U1D
4001BD_5V
U2A
4011BD_5V
U2B
4011BD_5V
U2C
4011BD_5V
U2D
4011BD_5V
R1
100Ω
R2
100Ω
R3
100Ω
R4
100Ω
VSS12V
R73kΩ
R83kΩ R10
9kΩ
R9
9kΩ
VSS12V
D1 D2
D3 D4
S3
Key = A S2
Key = S
VCC5V
Q11
IRF9530
Q12
IRF7404
Q7
IRF9530
Q6
IRF7404
R53kΩ
R63kΩ
R11
820Ω
Q2
2N2222A
Q4
2N2222A
R121MΩ
R13
9kΩ
R14
1MΩ
R16
9kΩ
Q1
2N2222A
Q2
2N2222A
Q3
2N2222A
Q4
2N2222A
U1A
4001BD_5V
U1B
4001BD_5V
U1C
4001BD_5V
U1D
4001BD_5V
U2A
4011BD_5V
U2B
4011BD_5V
U2C
4011BD_5V
U2D
4011BD_5V
R1
470Ω
R2
470Ω
R3
470Ω
R4
470Ω
VSS12V
R5820Ω
R6820Ω
R73kΩ
R83kΩ
R10
820Ω
R9
820Ω
Q5
MJF122GTIP122
Q6
MJF122GTIP122
Q7
MJF127GTIP127
Q8
MJF127GTIP127
VSS12V
R11
1.0kΩ
D1 D2
D3 D4
S3
Key = A S2
Key = S
VCC5V