team 1617: autonomous firefighting robot contest katherine drogalis, electrical engineering...
TRANSCRIPT
Team 1617: Autonomous Firefighting Robot Contest
Katherine Drogalis, Electrical EngineeringZachariah Sutton, Electrical Engineering
Chutian Zhang, Engineering Physics
Advisor: Professor John Ayers
Overview
• Project Overview & Contest Background• Mechanical Design & Layout• Sensors & Routing• Microcontroller• Flame Extinguishing• Power Supply• Budget
Design a Fully Autonomous Robot to Find & Extinguish a Flame
• Trinity International Robot Contest (April 1-3, 2016)• User initiated, autonomous start & navigation• Search for and extinguish burning candle• Design can be extended to real life situations
Trinity International Robot Contest
• 8x8’ plywood maze• Arbitrary start position• Competing in 2 of 3 levels• Timed trials• Unique robot• 31x31x27 cm robot
Level 1 Arena Level 2 Arena
Test Arena
Round Polycarbonate Body
• No rigid corners to bump walls• Electrical insulating property• Strong; Will not crack when cut• Threaded rod for support• Levels: Top to Bottom
o Start button; LED; mic; kill-power plug; handleo Flame detection sensors; extinguishero Microcontroller; laser scannero Driving motors; control circuit; batteries
Two Motors Independently Driving Two Wheels
• Can turn different angles simultaneously• Take commands from microcontroller• Option 1: Stepper Motors
o Position controlled: constant input voltage drives motor to specific positiono Draws current to maintain position - waste of battery power
• Option 2: Servomotorso Similar to Steppero Consumes power as rotates to position then rests - better, wastes less power!o Angle of rotation is limited to 180o (or so) back and fortho Complicated setup with PWM tuning
Best Option: DC Motor w/ Encoder
• Velocity controlled: constant input voltage drives motor to specific velocity• Can control position by applying velocity commands over a certain time
o Pulse-Width Modulation signal• FAST - 100 rpm• 12V - perfect for battery operation• Count wheel rotations with encoders• 64 counts per rotor revolution (6400 counts per wheel revolution)
Need to Sense: Walls/Obstacles & Flame• Range sensing options
○ Ultrasonic: cheap, easy to use, low interference, low resolution○ Infrared: cheap, range limited, interference prone, low resolution○ Laser: expensive, long range, low/no interference, processing required, high res
• Flame sensing options○ Look for presence of both light and heat○ Light: photoresistors/photodiodes, subject to external interference○ Heat: IR non-contact sensing, must work at range of ~1 m
Choice: 360o Laser Scanner by RoboPeak
• Scanner vs. Stationary○ Stationary: cheaper, would need to be mounted on scanning platform○ Scanner: set sample rate, configurable scan speed, built-in angular encoder
• Measurements in body reference frame polar coordinates (heading = 0º)○ “r” coordinate useful in finding wall discontinuities○ Need to convert to cartesian for SLAM
• 2000 samples per second○ Vary scan speed to control angular distance between samples○ Get ~1 sample per degree with 5.5 Hz scan rate
Video of Laser Range Sensor
Experimental Data
Flame Sensor
• RoBoard RM-G212 16X4 Thermal Array Sensoro produce a map of heat valueso able to pick up the difference 1.5m awayo low power consumptiono 16 x 4 = 64 pixelso FOV: 60º horizontal, 16.4º verticalo 0.02 Degree Celsius uncertainty
• Can find center of candle at close range• Have a particular pixel act as target location• May be unecessary to add light sensing
Experimental Setup
Experimental Flame Sensor Heat Map
Heat measurements at distance of 1.5 m Heat measurements at distance of 0.2 m
Candle in total field of view
Experimental Flame Sensor Heat Map
Routing/Navigation
• SLAM (Simultaneous Localization and Mapping)• Find current location in a map of landmarks
○ Use laser to pick out corners and terminal points in walls ○ Predict next position from current position and a given control command ○ Compare prediction with sensor result after command is executed○ Correct based on previous reliability of sensor measurements and predictions
• Adaptive
Routing/Navigation
Microcontroller
• Arduino Mega 2560o 256 kB of Flash Memoryo 8 kB of SRAMo 4 kB of EEPROM o 7 to 12 Voltso Highly versatileo Many available open source librarieso Programmable in C++
• Raspberry Pi (possible addition)o Helps the speed of processingo All real-time calculations with scanner data must be accomplished within 500 uso Will add if unable to make Arduino code this efficient
Flame Extinguishing
• Realistic: Compressed gas (CO2)o Best option for large-scale fire - bonus points!o Cartridge at the back of the roboto Extended nozzle at the front aligned with the sensorso Pointed directed at the candle flame
• Unrealistic: Fano Will make a large-scale fire worse!o Controlled by Arduinoo Fallback option
Power Supply and Other Requirements
• Rechargeable DC batterieso Two sets - use one while charging other - save time!o 4 separate cells - option to pull power from individual cellso Max 14.8 Vo 5500 mAho 532.2 grams
• Other requirementso Start button: green backgroundo LEDs: white backgroundo Microphone: blue backgroundo Kill-power plug: yellow backgroundo Handle
Budget
Questions?