Proceedings of the Multi-Disciplinary Senior Design Conference Page 3 s4

Proceedings of the Multi-Disciplinary Senior Design Conference Page 3

Project Number: P10217

Proceedings of the Multi-Disciplinary Senior Design Conference Page 3

ROBOT INTEGRATION AND TESTING

Patrick Arrigo / Mechanical Engineer (Team Lead) / Wes Coleman / Mechanical Engineer
Steven Guenther / Electrical Engineer / Adam Spirer / Electrical Engineer / Vernon Vantucci / Computer Engineer
Aaron Zimmerman / Mechanical Engineer

Proceedings of the Multi-Disciplinary Senior Design Conference Page 3

Abstract

The goal of this project is to build upon and otherwise enhance the Wandering Ambassador robotic platform as delivered by the Robot Locomotion (P10215) and Robot Navigation (P10216) teams. The overall goal of the Wandering Ambassador track is to create a robot with the goal of autonomously maintaining a living plant (ivy) while wandering the grounds of RIT through the use of several positional and environmental sensors. The motivation for the project is the creation of an innovative and, hopefully, interesting "ambassador" that showcases RIT's commitment to innovation and the "green" initiative. The general goals for robot operation include a full day of autonomous operation wherein the robot will determine, through use of the sensors mentioned above, how best to address the needs of its plant companion (i.e. sunlight and water), as well as generating pedestrian interest in the robot through motion, all the while operating in a safe manner for both itself and anyone nearby.

This team is working in conjunction with the Robot Applications team, P10218, who are developing advanced software features such as interfaces between the robot and internet users through both social networking sites such as Twitter and Facebook as well as more advanced control and feedback via a specialized web page.

Nomenclature

ADC - Analog to Digital Conversion

I2C - Inter-Integrated Circuit

MSP430 - a microcontroller made by Texas

Instruments

PWM - Pulse Width Modulation

RIT - Rochester Institute of Technology

TI - Texas Instruments

F.R.E.D.- Free-Roaming Environmental Drone

CAD- Computer Aided Design

PCB – Printed Circuit Board

introduction

The field of robotics is one of the faster-growing areas in technological advancement. From building automobiles to performing remote surgeries to disarming explosive devices, robotics provide the ability to perform jobs more economically, more precisely, and more safely than humans can do the same task.

The grant (RIT Provost's Learning Grant) provided for funding the project was provided in order to create a robotic platform that not only showcases efforts in autonomously locomotive robotics, but that also shows efforts to address the "green" ideology prevalent in modern industry.

To those ends, F.R.E.D. (as he has become known), is being developed to autonomously navigate areas of campus (using a combination of on-board software married to various sonar and infrared sensors) while addressing the biological needs of a plant affixed to his back. Correct light and soil moisture parameters for the plant are loaded into F.R.E.D.'s software, and in conjunction with light and humidity sensors will prompt activation of F.R.E.D.'s on-board watering system and/or navigation towards or away from sunlight.

Navigational commands are sent to two electric motors and chain-driven wheels to provide navigation in all directions.

Currently, F.R.E.D.'s systems are powered by a 12-volt deep-cycle marine battery with the foundations being laid for the integration of renewable sources of energy (i.e. sunlight) in future iterations.

F.R.E.D.'s shell is constructed with layers of fiberglass in accordance with a design generated and passed down by the Robot Locomotion team (P10215). F.R.E.D.'s fiberglass shell was painted with a design that highlights his green purpose- vines, clouds, and rain.

In the event of a mechanical failure or other servicing needs, F.R.E.D. has a manual return option via two handles and spring-loaded casters through which he can be moved as required.

design methodology

Figure 1: 3D Model Rendering of Robot

Electrical

The robot’s sensors are interfaced using two TI MSP430 microcontrollers, which provide low-level processing functions to assist the robot’s BeagleBoard main computer. Among the sensors for navigation include 4 sonar sensors for distance measurement and 3 infrared sensors for ground sensing. There is also a digital compass and accelerometer. Among the sensors for locomotion are two optical encoders to detect motor revolutions. Among the sensors for plant care are a photoresistor for light level, humidity sensor for determining the soil condition, and a temperature sensor for reading ambient temperature. This interface was put together by previous team work. For each of these MSP microcontrollers, the sensors are read individually in a cyclic executive loop, keeping the MSP’s registers updated constantly with new data. This data can be read via the I2C interface on the BeagleBoard at regular intervals.

A third MSP430 microcontroller is present for operating the robot’s two drive motors, which are controlled via four half H-bridges with a varying PWM signal to determine speed and direction. A standard serial interface over RS232 with the BeagleBoard allows control of these PWM signals. In addition, the MSP430 interfaces with two optical encoders (one for each motor) to keep track of how far the robot has traveled.

Figure 2: Board Hierarchy

A primary goal was to verify functionality of this interface put in place by previous team work and test it to determine reliability. Specifically, integration of these sensors required a redesign of the physical interface between the MSP430 processors and the sensors themselves. For analog sensor readings, the MSP430 and the sensor voltage dividers were recalibrated to operate on a 3.3 V reference supply instead of the original 2.5 V supply. In addition, firmware was redeveloped for a new accelerometer that was ordered after the original one was damaged. Verification tests were performed using real-time debugging to confirm reliable operation of the cyclic executive loop, and correctness of sensor data was verified as well.

The original optical encoders were replaced with new ones after the original ones were determined unreliable. These encoders were interfaced using the on-board MSP 12-bit ADC. Firmware was developed and tested for these new encoders, and the entire serial interface for motor control has been continuously extended to implement new features; two new features include the ability to reset the optical encoder data as well as a feature to correct steering via a PWM offset value. The need for the latter addition was considered as a means to provide calibration ability, as load imbalance and other mechanical issues may cause the motors to have slightly different speeds even on the same PWM signal.

Shell

Shell Construction

From P10216 (Locomotion), the Integration team inherited a complete single-piece CAD model for the robot’s decorative and protective shell, and an incomplete manufacturing plan for said shell. A curvature analysis was performed on the CAD model to aid in deciding on a fabrication method for the shell.

Figure 4: 3D Model Rendering of Robot
cool colors=slight curve, hot colors=aggressive curve

Due to the aggressively curved shape of the shell, the concept selection process resulted in a decision to fabricate the one-off shell out of fiberglass. The shell model was split into four distinct parts to ensure a favorable draft angle for each piece.

From the CAD model, profiles of each piece were printed, one for each Cartesian dimension. The profiles were then transferred onto a block of foam which, through hot wire cutting and manual sculpting, became the positive molds from which the fiberglass panels were made. Joint compound was used to fill any gaps left by cutting, and sharp corners were rounded to improve the mold’s ability to hold fiberglass.

Figure 3: 3D Model Rendering of Robot

First, a 1:3 scale model of the largest part was fabricated using three layers of 3 oz fiberglass cloth. The 1:3 scale model proved that it was possible to remove the intact foam positive from the fiberglass lay-up once it cured if the mold is covered in epoxy-impervious plastic. This allows the positives to be reused if spare parts are needed in the future.

The full sized parts were fabricated using a wet lay-up process with four layers of 10 oz fiberglass cloth and 30 minute epoxy resin, mixed by weight. In this process, a sheet of fiberglass cloth is draped over the plastic covered positive mold, then coated in mixed epoxy resin. This layering process is repeated until the requisite number of layers is attained. The fiberglass material was selected for its combination of thickness and flexibility in curves. The epoxy was selected for its long working time and ability to cure at room temperature, as an autoclave large enough for this project was unavailable. The wet lay-up on positive mold process was selected for simplicity and mold reusability.

Although the fiberglass shell pieces required trimming and had very loose tolerances and rough surface finish, the fabrication process produced a very near net shape product flexible enough to force onto the robot’s chassis. Moreover, the reusable molds make the process repeatable.

Electrical System

The electrical systems of the robot were in a dire state when handled off from the previous two teams; the robot was nonfunctional in every way. The entire system needed to be reworked, in the few cases where the original design could be used, it was. The electrical systems were broken down into six subsystem, power, motor control, h-bridge, navigation, plant, and BeagleBoard. Due to the lack of accurate or up to date schematics from previous teams, each subsystem was deconstructed from the original boards and redesigns integrated to meet specific customer specifications.

Figure 5: Electrical boards as received

Three phases of electrical development were conceived. The first phase was to construct a working prototype of the complete system. Phase two was to produce a custom PCB to encapsulate stage one work but still include development boards and other working circuitry from the previous teams. Phase three would be to develop a PCB with all custom hardware, reducing all development board to their necessary circuits and placing them directly on a unified design. Phase three is intended to be carried out by subsequent teams.

Figure 6: Phase One Prototype Boards

Phase one's goals were simple, produce a reliable working robot with an accurate unified electrical schematic. A standard ‘perf board’ was used to create the new design. The motor controller board and the H-bridge from the previous locomotion team were carried over; the only modifications needed to it were the inclusion of the new wheel encoders. A new power system was designed using simple 5 volt and 3.3 volt regulators. Navigation and plant systems were created using the previous teams boards as reference. Issues corrected from the previous teams include but were not limited to, incorrect voltage levels, poor soldering, improper connections, and parts and traces that did not meet customer specifications or electrical specifications.

Phase two was to produce a clean concise PCB using the complete schematics produced in phase one. The development boards, which included the BeagleBoard, the MSPs, motor controller, and H-bridge, were seated on the PCB and sockets or jumper connections were used to bring the relevant signals down to the PCB. The idea was that these devices were working and reliable as individual units, so they should be brought into the new system as such. The new power design and all interconnects between the subsystems were integrated onto the PCB.

Phase three would be to reduce the development boards (BeagleBoard, MSPs, motor controller, and H-bridge) down to their necessary components and included them on one fully integrated PCB. Having a specialized integrated PCB would reduce the costs of manufacturing multiple robots for future use. Future teams would complete phase three.

TESTING

Testing of specific customer specifications was not completed due to time constraints and setbacks. A Testing Plan was developed, however, and was laid out in the following manner; preliminary tests consisting of proper initialization of FRED (i.e. powering on with no detected faults), were to be ran first. Following the preliminary tests, basic operation consisting of testing customer specific requirements for stopping distance, turning radius, rotational velocity, translational velocity etc. would be run. After initial testing is completed, complex/combined movement of FRED was to be tested (combined translational and rotational movement). Following full manual operation of the robot, the autonomous functionality of FRED, such as obstacle detection, fault monitoring, plant monitoring and care etc. would be tested. Upon completion of all previous tests, full duty cycle runs would be done to ensure full functionality and on-site performance. To ensure adequate overall performance, the higher level tests, such as complex movement and duty cycle runs, were to be completed at various locations on the campus consisting of, but not limited to, the Gordon Field House, Simone Circle, the Kodak Quad and the 4th Floor Senior Design Center.

Proof-of-design testing was carried out at the ImagineRIT event. FRED operated continuously with only minor interruptions in remote control operation for approximately 7 hours, with a total battery voltage drop of approximately 0.3V (12.2V to 11.9V as measured after the event). The robot was successful in generating the interest of passers by- one of the initial goals of the project.

Acknowledgments

Prof. George Slack (Guide)

Dr. Ferat Sahin

Dr. Pratapa Reddy

Dr. Hany Ghoneim

Prof. John Wellin

RIT Aero Club

Joellen Zimmerman

RIT Provost's Learning Grant