IMDL Fall Final Report

Transcription

1 IMDL Fall 2014 Final Report Designer: Jacob Easterling Robot Name: Clean Sweep Course Number: EEL 4665 Instructors: Dr. Arroyo Dr. Schwartz Dr. Diaz Teaching Assistants: Andy Gray Nick Cox

2 C l e a n S w e e p 2 Table of Contents Abstract...3 Executive Summary.4 Introduction Purpose. 5 Overview. 5 Preview. 5 The Mat Layout Sectors....6 Seeker Bot Integrated System...7 Mobile Platform...8 Actuation Sensors. 9 Behaviors..10 Sentry Towers Integrated System...11 Stationary Platform...12 Sensors...12 Behaviors...13 Results...13 Debris Bot Integrated System...14 Mobile Platform...15 Actuation...15 Sensors...15 Behaviors...16 Limitations...16 Raspberry Pi Why Use a Pi?...16 Triangulation Algorithm...16 Limitations...16 Conclusion Work Accomplished...17 Limitations...17 Improvements Appendices Sentry Tower Schematic.18 Source Code...19

3 C l e a n S w e e p 3 Abstract The purpose of my project was to design and develop a team of robots that could work together clear objects out of a known perimeter. In order to make this possible I identified three elements which each robot needed: situational awareness, communication, and task management. Before we dive into these points let us first identify the task to complete. The task I decided to take on was to use team work to identify and clear all objects out a known area. Situational Awareness This block proposes the questions of where am I? and where are my teammates? In order to answer these questions I developed a system of three sentry towers which can time stamp a high frequency (roughly 8 khz) when prompted. The signal is broadcasted by the Seeker Bot (a small robot equipped with an amplified speaker) who will be traversing the known perimeter looking for objects. Once locked on the signal, each tower can relay the distance in-between the Tower and the Seeker Bot to an independent processor (raspberry pi) which will generate an (x,y) coordinates for the object. Communication As is true in any society, communication is the fundamental in order for a team to accomplish any meaningful task. Communication is also foundational in order to orchestrate a team of intelligent machines. To address this each of my robots is equipped with an XBEE wireless Transmitter/Receiver with a unique name (Tower1, Seeker Bot etc.) to allow for a dynamic conversation. Task Management This part might seem intuitive but I have found it is absolutely essential to define what services each robot will provide. Here is an overview of what each team member will contribute the project. Seeker Bot o Moves inside known perimeter looking for obstacles o Broadcasts Homing Signal to Sentry Towers when an object is found Sentry Towers o Time stamps the Homing Signal from Seeker Bot o Sends distance in-between Tower and Seeker Bot to Raspberry Pi. Debris Bot o Sweeps obstacles out of the perimeter Raspberry Pi o Manages all communication on the field o Processes data from Sentry Towers to locate object within in the perimeter

4 C l e a n S w e e p 4 Executive Summary As the world of robotics continues to grow in complexity we are beginning to see a shift towards the coordination of multiple robots to perform dynamic tasks. This behavior, also referred to as Swarming was the inspiration for my project which uses five robots to identify, locate, and clear objects out of a known perimeter. In this summary we will look over the details of the components that make up my robot team. The Mat In order to demonstrate my team s abilities I have designed a 3 x5 Mat which acts as a playing field for the team. Along the perimeter of the mat are tick marks which allow the Debris Bot (see below) to move towards an objects position when located. Sentry Towers These towers are used to assist in the object location process by time stamping a series of high frequency pings sent from the Seeker Bot (see below). These towers are placed on three of the four corners of the mat. Seeker Bot This robot is a small lightweight mobile platform which moves along the inside of the mat looking for objects. When the Seeker Bot locks onto an object he will ping the Sentry Towers both sound and RF to correlate his position. Once his position is identified the Seeker Bot will search for the nearest perimeter line and wait until the object has been cleared by the Debris Bot. Once the Seeker Bot receives word that the object has been cleared he will resume his search pattern. Debris Bot The Debris bot is a heavy duty aluminum/steel mobile platform designed to clear objects out of the mat. While the Seeker Bot is searching for objects, the Debris Bot enters sleep mode in the corner of the mat while waiting for instructions. Once woken up, the Debris Bot will receive the coordinates for the object to clear and will move up the mat along the perimeter line. As he traverses the mat, he will keep track of his position by counting the tick marks along the perimeter line. Once he has reached the objects position he will sweep across the mat and push the object out of the perimeter. He then will return to the corner and enter back into sleep mode. Raspberry Pi This processor manages all the RF communication of the team. He also reads in sensor data from the Sentry Towers and triangulates the position of the Seeker Bot when prompted.

5 C l e a n S w e e p 5 Introduction Purpose The purpose of this project is to assemble a team of low cost simple robots that can work together to complete a task too complicated for any one of the robots to complete on their own. Objective The objective for the Clean Sweep robot team (see Fig 1.) is to identify, locate, and remove objects within a known perimeter utilizing the advanced Preview In this Report you will find details on each of the members of the Clean Sweep robot team including their integrated systems, platform design, actuation, sensor usage, behaviors, and results. Fig. 1 A quick look at the members of team Clean Sweep

6 C l e a n S w e e p 6 The Mat Fig. 2 This shows how the mat is divided into virtual sectors Layout In order to demonstrate my team s abilities I have designed a 3 x5 Mat which acts as a playing field for the team. Along the perimeter of the mat are tick marks which allow the Debris Bot to move towards an objects position when located. Sectors In order to accommodate for errors in the triangulation process, I have divided the mat into 3 virtual sectors (se Fig. 2). Within each sector there is a single object which the Seeker Bot will discover.

7 C l e a n S w e e p 7 Seeker Bot Integrated System The Seeker Bot uses two MSP430 microcontrollers to operate the sensors, motors, and communication. Ultrasonic and IR range finders enable the Seeker Bot to identify obstacles while an array of three IR line reader forces the robot to search only the area of the mat. The block diagram below (Fig. 3) shows how each component is connected on the Seeker Bot. IR Rangefinder IR Line Reader x3 MSP430 G2553 IR Rangefinder 6V DC Motor Motor Controller 6V DC Motor XBEE S1 MSP430 G2553 Ultrasonic Rangefinder Speaker Array Fig. 3

8 C l e a n S w e e p 8 Mobile Platform For my design, my intention was to have the Seeker Bot be placed on a small platform. For this reason I chose to purchase the off the shelf Zumo Chassis (see Fig 4.) Despite the ready to run appeal, there were a few modifications to the chassis that had to be completed before the Robot could run as needed. First I needed to file down the plastic around the spokes of the wheels as it was grabbing the tracks and causing the motors to seize. Secondly I added additional levels to the base platform to fit the onboard hardware. The first level accommodates all the microcontrollers, Xbee and some select sensors (see Fig. 5) while the second level holds an amplified speaker which be used to work with the Sentry Towers. As seen in Figure 6, I also mounted a parabolic dish above the amplified speaker to aid in the spread of the signal when broadcasting. Fig. 4 Actuation The Seeker bot uses two 6V 100:1 geared motors to maneuver around the mat (see Fig. 7). Motor Specifications o Nominal Voltage: 6V o Free RPM: 315 o Stall Torque: 25 oz-in (1.8 kg-cm) o Stall Current: 360mA o Reduction: 100:1 o Size (mm): 1.42" x 0.39" x 0.47" o Weight: 0.35oz / 10g Fig. 5 Fig. 6 Fig. 7 Each motor is run through an H-Bridge motor controller which modulates the input voltage.

9 C l e a n S w e e p 9 Sensors Ultrasonic Range Finder (Fig. 8) o Vendor: Banana Robotics Model Bumber HCSR04 o Specs Input Voltage 5V PWM freq: 40Hz Range: 2cm to 40m 15mA avg consumption o Application Used to locate objects on mat See Appendix for source code IR Range Finder (x2) (Fig. 9) o Vendor: Robot Shop Model Number GP2Y0A21 o Specs Input Voltage: 5V Range: 10cm to 80cm 39ms response time 40mA avg consumption o Application Used for moving around obstacles IR Line Reader (Fig. 10) o Vendor: Spark Fun Model Number QRE1113 o Specs Input Voltage 3.3V 25mA supply current Optimal sensing distance: 0.125" (3mm) o Application Used to detect the mat perimeter line Fig. 8 Fig. 9 Fig. 10

10 C l e a n S w e e p 10 Behavior The Seeker Bot s job is to seek out objects on the mat and relay the object s position to the Sentry Towers. Starting in the center of the mat, the Seeker Bot will run an initial calibration where he will send out nine 6 khz short bursts using a high precision timing algorithm (see Fig 11 for state machine). Once calibration is complete, The He will move around the Mat at random searching for objects using his rangefinders. Once an object is detected, the Seeker Bot will run through the subroutine shown in Figure 11 and then inform the Raspberry Pi that new data is ready to receive. The Seeker Bot will then wait for confirmation from the Pi to either continue or ping the towers again. Once the location is confirmed the Seeker Bot will head towards the nearest perimeter line and waits while the Debris Bot is clearing the object. If a collision between the two robots is eminent, the Seeker Bot can receive commands to relocate to a new line while the Debris Bot is on the Mat. After receiving confirmation that the object has been cleared, the Seeker Bot will resume his normal search pattern. A Exit Subroutine 0 Cnt < 9? 1 Send RF Trigger to 3 towers A Delay 250ms Enable Speaker to Ping Cnt++ Delay 250ms Read in from Towers Fig. 11 Ping Subroutine 0 1 All Towers Reported?

11 C l e a n S w e e p 11 Sentry Towers (Special System) Integrated System Each Sentry Tower uses an MSP430 microcontroller to monitor sensor inputs, and communication. There are also a number of IC chips which condition signals coming into the MSP including a High Pass Filter, and V/U Meter. Below is a block diagram of circuits on operating inside each Seeker Tower. In addition to the block diagram, there is also a full schematic for the Sentry Towers in the Appendix. LM386 Audio Amplifier LM386 Audio Amplifier LM741 40dB High Pass Filter LM741 40dB High Pass Filter LM3915 V/U Meter LM3915 V/U Meter XBEE S1 MSP430 G2553 Fig. 11

12 C l e a n S w e e p 12 Stationary Platform As shown in Fig. 14, the Sentry Towers are design around natural listening systems which includes two ears and a brain. Each ear is mounted to a rotating platform connected to the central stepper motor (note: the stepper motor is no longer in use with my design) (see Fig. 14). Each ear can be independently rotated in both the x and y directions to allow the operator to determine the optimal angle between the two ears. In order to lower false signal locks, I also added cones to fit over the mics to narrow their field of listening. The base of the platform also has an LED bank built in (see Fig. 13) which provides visual feedback during communication. Also seen in Figure 13 is that the Sentry Tower sits on top of a power supply which produces a clean 10.5V to the filtering circuits and 3.3V to the MSP430. Fig. 13 Sensors Unidirectional Mic (Fig. 15) o Vendor: Digikey Model number: CMI-5247TF-K o Application The mics in my Sentry Towers are used to determine the source of the Seeker Bot s 6 khz homing signal. o Specs Input Voltage 1.5 ~ 12V Frequency Range: 70Hz ~ 20kHz Impedance: 680 Ohm 500μA avg. consumption

13 C l e a n S w e e p 13 Behavior While the Seeker Bot is searching for objects, the Sentry Towers sit idle, poling a start signal from the Seeker Bot. Once an object has been found by the Seeker Bot, he will send the start signal simultaneously to all three towers which will tell them to enable their time-stamp program. 250ms later the Towers will each receive a 6 khz pulse from the Seeker Bot. Once the first pulse has been received, they time-stamp the difference in time between the RF trigger and the 6 khz pulse. Each Tower then sends confirmation back to the Seeker Bot that they have received their data and are ready to receive. Once all the Towers have sent confirmation back to the Seeker Bot, he will send the second of nine pulses. After this process, the Sentry Towers then holds their data until called upon by the raspberry pi (see later section). Results Signal Conditioning o As shown in the image below, you can see the effects of the high pass filter on incoming signals 1 khz 5 khz 10 khz Accuracy o Given errors produced by hand-crafting both the boards as well as the microphone shields, there is roughly a 15-20% margin of error when pinpointing the position of an object using sound. To combat this I have split the mat (as mentioned previously) into the sectors with a single object placed inside each. By searching for sectors over actual positions, I have narrowed the margin of error to under 10%.

14 C l e a n S w e e p 14 Debris Bot Integrated System Seeker Bot utilizes four MSP430s to handle all operation on the platform. Ultrasonic range finders enable the Debris Bot to identify obstacles in its path while an array of IR line reader will allow the robot to consistently find an object s location on the mat. The block diagram below (Fig. 16) shows how each component is connected on the Debris Bot. IR Line Reader x3 MSP430 G2553 Ultrasonic Range Finder x2 Motor Controller MSP430 G2553 Motor Controller XBEE S1 MSP430 G2553 IR Line Reader MSP430 G2553 Fig. 16

15 C l e a n S w e e p 15 Mobile Platform In order to be coherent with my design plans, I needed the Debris Bot to have a platform which was larger than the Seeker Bot. For this reason I chose to purchase the off the shelf DF Robot 4WD Mobile Platform (see Fig 17). This chassis is roughly 2.5x the size of the Seeker Bot and can easily fit 12 sensors and custom PCB designs required to make this robot operational. Actuation The Seeker bot uses four 6V 100:1 geared motors to maneuver around the mat (see Fig. 18). Reason for purchase o These were the motors which were designed to work with my Chassis. Motor Specifications o Operating Voltage Range: 3~7.5V o Rated Voltage: 6V o Max. No-load Current(6V): 170 ma o No-load Speed(6V): 160 rpm o Max. Output Torque: 0.8 kgf.cm o Max. Stall Current: 2.8 A Each motor is run through an H-Bridge motor controller which modulates the input voltage. Sensors Ultrasonic Range Finder (x2) o Vendor: Banana Robotics Model Number HCSR04 o Specs Input Voltage 5V PWM freq: 40Hz Range: 2cm to 40m 15mA avg consumption o Application Used to locate objects on mat See Appendix for source code IR Line Readers (x4) (Fig. 19) o Vendor: Spark Fun Model Number QRE1113 o Specs Input Voltage 3.3V 25mA supply current Optimal sensing distance: 0.125" (3mm) o Application Used to detect the mat perimeter line Fig. 17 Fig. 18 Fig. 19

16 C l e a n S w e e p 16 Behavior The Debris bot is a heavy duty aluminum/steel mobile platform designed to clear objects out of the mat. While the Seeker Bot is searching for objects, the Debris Bot enters sleep mode in the corner while waiting for instructions. Once woken up, the Debris Bot will receive the coordinates for the object to clear and will move up the mat along the perimeter line. As he traverses the mat, he keeps track of his position by counting the tick marks along the perimeter line (see Fig. 20). Once he has reached the objects position he will sweep across the mat and push the object out of the perimeter. He then returns back to the perimeter line and aligns himself with it and returns to the Fig. 20 corner. Once docked, the Debris Bot will notify the raspberry pi that the sector is Example cleared and return to sleep mode. perimeter line Limitations Due to the crudeness of IR sensors, I have had some difficulty with the re-aligning of the Debris Bot once off the line. On occasion, the robot requires manual repositioning. Raspberry Pi For my Clean Sweep Team I use a Raspberry Pi computer to manage all wireless communication, and handle the advanced triangulation algorithm. Why Use a Pi? Raspberry Pi s have a distinct advantage over MSP430s in that they have floating point hardware on board. This allows me to perform the difficult triangulation algorithms in a fast and efficient manner. Triangulation Algorithm The Algorithm that I have created exercises Heron s Formula (shown below) which allows you to determine the height and distance to the center of a triangle given only the side-lengths. Given that the distances between each Sentry Tower are fixed, we can determine the position of any object on the mat. a + b + c p =, p = 0.5 perimeter of triange; a, b, c = side lengths 2 A = p (p a) (p b) (p c) h = 2 A, h = height of triangle b x = c 2 b 2, x = distance to center of triangle Limitations Since there are errors in the distances reported by the Sentry Towers, there is a chance that the reported values do not produce a valid triangle. This error has been corrected in software by telling the Seeker Bot to look for a different object if this issue arises. He will then come back at a later point and try again.

17 C l e a n S w e e p 17 Conclusion Conclusion Work Accomplished In summary of my work, I have successfully built and coordinated a team of five robots that work together to identify, locate, and clear objects out of a known perimeter. Limitations As noted in previous sections, my Sentry Towers suffer from inaccuracies due to error in the handcrafting of their design. Also the Debris Bot has some difficulty remaining on the line while progressing towards an objects location. Improvements If I could do this project over again, I would re-design my Sentry Towers to use high-precision parabolic dishes to hone in on a sound in replacement of the shielding I currently use. Clean Sweep Team

18 C l e a n S w e e p 18 Appendix Sentry Tower Schematic

19 C l e a n S w e e p 19 Source Code All source code can be found at this web address: