Project E.A.S.I II. POWER SUBSYSTEM.

Project E.A.S.I Heath Cissell, Stephen Miles, Patrick Shiver, Hung Tran Dept. of Electrical and Computer Engineering, University of Central Florida, Orlando, Florida, 32816-2450 Abstract Project E.A.S.I or Electronic Assistant for the Seeing Impaired is device that will replace the whitecane. This device is smaller and more discreet than a cane providing tactile and audio feedback in response to ultrasonic sensors that measure the area in front of the user. The devices currently available to the visually impaired community are limited with most developed several decades ago. E.A.S.I is designed to give users more options not by simply enhancing current options but by replacing them with modern day sensors and applications. Index Terms Seeing Impaired Aid, object detection, wireless communication, application software, tactile feedback, mobile communication. I. INTRODUCTION. The visually impaired across the globe face many difficulties that sadly most of us take for granted. As of 2013 an estimated 7.3 million people are reported to have a visual disability. Each of them must struggle with their independence within this world, often having to rely on others to simply get from point A to point B. Existing technologies have not changed much from their invention. The two most common forms of assistance available are Seeing Eye dogs, commonly called guide dogs, and a walking cane commonly called a white cane. Both systems are effective in keeping the user from walking into things and getting around by using these tools to sense their environment. But each has their own advantages and limitations. In the past people have tried to improve upon these systems by placing sensors or systems onto the existing cane or service animal. Our design is a handheld device that can be held discreetly using distance finding technology to replace the need for a walking cane and help to restore a small bit of independence while being more discreet. A. Constraints E.A.S.I is at its core a safety device which requires a high level of functionality. Based on its environment and usage, our device had to be lightweight, small, a high battery capacity and accurate distance sensing and while being able to provide tactile feedback. As it stands, the E.A.S.I. is designed to weigh less than 2.2 kilograms, fit into someone s hand, operates for at least 12 hours, and is accurate to a range of 2 meters with a field of view of roughly 90 degrees. B. Design In accordance with our design constraints, we first identified each subsystem and group member s responsibility. Trying to keep the design as small as possible was difficult. Block Diagram Fig. 1. Battery Charging System Battery Tactile System Packaging Bluetooth System Controller Distance Sensor Systems Group 35 Heath Stephen Patrick Hung Block Diagram of E.A.S.I. II. POWER SUBSYSTEM. Android App Audio System Voice Sensing Under the constraints of being mobile and small but providing tactile feedback made the selection of an appropriate battery difficult. The mechanical actuators require a great deal of power especially initially to trigger the change but far less to hold the actuator on. The main purpose of the tactile circuit is trying to reduce the power consumption of the actuators. However, to meet a daily use of 12 hours required at least a 6000mAh battery. The power subsystem provides 3 voltage buses to all the components in the circuit. The voltage buses include two 5V buses and one 3.3V bus which are supplied by 3 buck switching regulators. Due to the nature of the mechanical actuators, the power subsystem focuses on efficiency. The linear regulators though bigger in size offered the greatest efficiency at roughly 2 percent. Finally, due to safety and efficiency we utilized the MCP73213 battery manager for regulating the battery charging. This provides a steady voltage along with a programable current that can control the rate of battery charge. The battery manager also provides the status of the charge giving indication of when the battery is fully charged. 1

A. Lithium-Ion Polymer Battery We initially went with a Li-Ion battery but due to size and cost we decided to go with a Lithium polymer battery. In going for efficiency, the charging voltage for this battery is 8.4V with a current of 1A. [1] These parameters allow for 100 percent capacity at full saturation in just over 6 hours. We could charge this battery faster using a higher current rating however that would result in less capacity which would impact our daily run time. The lithium polymer battery was subjected to a series of tests to show battery life. We found that the battery life is just over 12 hours with cycling of only 4 actuators. Though theoretically our device could draw more power this is a balanced test of regular use. Shown below is the Tenergy Battery pack s voltage shown vs time. Fig. 2. LiPo battery voltage test over time. III. DISTANCE SENSING SYSTEM. One of the primary systems of our project includes the sensing elements. The sensing element system provides all the inputs to the microcontroller. This system is vital to our project for it provides the eyes for our device and for the user. The system is set up into an array of sensors for both redundancy checking and to serve 2 purposes, to sense the distance in front of the device and detecting objects in the immediate field of view. A. Ultrasonic Sensor The group elected to go with three MAXSONAR LV1 sensors. These sensors are designed to generate high frequency sound waves and receive the echo reflected by the target. These sensors are used in a wide range of applications and are very useful when it is not important to detect colors, surface texture, or transparency. This sensor offers several possible ways to communicate with the microcontroller unlike the HC-SR04 such as RS232, Analog Voltage, Pulse Width and TTL serial. These robust options will give us different implementation options should the need arise. The main advantage of this sensor besides the increased features is the size. At half the size of the HC-SR04 this sensor was well worth the cost. This sensor also used less power than its larger predecessor. B. Performance Analysis There are several factors that can affect the accuracy of an ultrasonic sensor and though accuracy is not a major concern for our design it is important to understand that they are present. Temperature and the reflection angle of the reflection surface are the biggest factors when determining accuracy. However, due to the nature of our design in using an Array of 3 sensors to sense the linear distance and objects to the left and right of the user, CrossTalk has been a major issue for us. Crosstalk is an issue that occurs when multiple ultrasonic sensors or proximity of objects can degrade the signal accuracy by inducing extra noise into the system. In our original design, we used a servo motor to sweep a single sensor back and forth much like a radar system but it produced such inaccurate results we decided to use 3 fixed sensors. Though it degraded the accuracy of sensing a full 3d field of view it increased the accuracy of the data we could process. The chart below shows the analysis of the sensing data for both the linear system as the choice between MaxSonar LV1 and LV4 was needed. MaxSonar LV4 has a narrow beam which we thought was perfect for the linear component of our device. However, upon testing we discovered that the narrow beam could not pick up major objects such as people past the 5-foot mark. For safety, our device requires at least 2m and ideal 3m of detection distance which MaxSonar LV1 supplied. Due to this we utilized this sensor 3 times within our sensor array. MaxSonar LV4 MaxSonar LV1 Actual Detected Detection Detected Detection Object Distance Distance Spread Distance Spread 2x2 Vertical 132 134 132 36 2x2 Vertical 108 107 107 36 2x2 Vertical 84 83 83 48 2x2 Vertical 60 60 12 59 36 2x2 Vertical 48 48 8 48 36 2x2 Vertical 36 36 12 35 24 2x2 Vertical 24 23 12 23 24 2x2 Vertical 12 12 6 11 10 Fig. 3. MaxSonar LV1 and LV2 Comparison. 2

IV. TACTILE FEEDBACK SUBSYSTEM One of the other major subsystems is the system of peripheral components which will be responsible for relaying all communicable information to the end user via touch. These peripherals consist of seven linear actuators and two haptic vibration motors. Two of the linear actuators will be used exclusively for GPS signaling, three will be used for object linear distance signaling, and two more will be used for object position indication. A. Vibration Motors There is an enormous variety of vibration motors, so given the plug-and-play flexibility of the haptic IC we are using, we felt it more prudent to order several varieties and subjectively test them and see which felt the best. We looked to minimize power consumption while keeping a large enough output force so allow for more easily distinguishable patterns. We found that the Model 308-104.001 from Precision Microdrives fit our needs nicely. B. Haptic IC We decided to drive our vibration motors using an I2C dedicated controller in order to offload complex PWM waveform generation from our MCU and allow for the capability of simple drop in replacement testing for several different varieties of motor. We ended up going with TI s DRV2605, in large part due to its easily accessible ROM waveform library. The one drawback of this chip is that it s I2C address is nonconfigurable, meaning we require an I2C multiplexer to use two. We used the TCA9548A for this task. As far as implementing these chips, most circuit design involves simply connecting the chips I/O pins as specified in their datasheets, and determining the range of acceptable pullup resistance values for our I2C bus. This process is detailed in the I2C standard [2]. The minimum resistance is as follows, where V OL and I OL are the minimum logic low voltage and current levels: R min = (V CC V OL )/I OL V OL and typical I OL can be found in the I2C specification, in Table 9. Using these values, we find that R min = 725Ω 1kΩ. Our goal is ultimately to maximize the pullup resistance without increasing the signal rise time past what is allowed by the I2C bus speed, to minimize overall current consumption. The maximum pullup value relies on an estimation of our I2C bus capacitance, which is based on the total trace length and the capacitance of each chip in series along the bus: R max = t r /(0.8473 C b ) Where t r is the maximum allowed rise time. Referring to the datasheets for the two I2C slave devices shows a bus capacitance of roughly 20nF per device. Given that two chips will be active at any given time, and a t r = 1000ns for standard mode [2], R max 30kΩ. Thus, we chose pullup values of 20kΩ to allow for any excess parasitic capacitance in the traces. C. Linear Actuators We chose the ZH0-0420S-05A4 small scale 5V push solenoid for our linear actuator, as it was one of the only push variants available at that size and voltage rating. To control these solenoids, we will rely on a power N-FET driven by GPIO from our MCU. Even at this small scale, solenoids typically consume quite a lot of power and get quite hot if left actuated for long, so most of the design effort that went into this subsystem was directed towards finding ways to minimize this power consumption as much as possible. To do this, we took advantage of a few key principles; it takes much more current to initially actuate a solenoid than it does to keep it actuated. Therefore, we only need draw high current for the fractions of a second it takes to move the solenoid plunger. Additionally, though the solenoids are rated to operate at 1.1A (5V, 4.7Ω ESR), they will still actuate with much less current than that, just with less force. Since our application only calls for signaling and not actually performing work, this reduced force is irrelevant. To implement the first principle practically, we first tested to determine exactly what the minimum necessary DC voltage was that would A) move the plunger to its actuated position, and B) keep it in position given that it is already actuated. The table below shows the results of this test: Voltage (V) Current (ma) A) Actuation 2.3 489 B) Holding 0.9 191 Fig. 4. Measured actuator characteristics This observation is what informed our decision to include two 5V regulators: each one is rated for 1A, and our actuator configuration will worst case have four solenoids actuated at a given moment. This means 3

that worst case current draw will be 1.96A, though only for fractions of a second at a time. We attempted two methods for reducing the actuation current from 1.1A. The first and simplest is to simply add a series resistance in line with the solenoid load, which will drop the DC voltage across the solenoid down to the measured 2.3V. The downside to this approach is that a fairly substantial amount of power will inevitably be dissipated through this resistor as heat. Due to its simplicity, we implemented this solution as a backup. The second method avoids this waste by taking advantage of the actuators inherent inductance. By pulsing the input voltage at a high enough frequency, the current draw won t have time to catch up to the change in voltage. As frequency increases, the instantaneous current draw will approach DC, and its magnitude will decrease as the current has less time to rise during the positive voltage cycle. By experimentation, we found a 20kHz, 60% duty cycle pulse adequate for supplying enough current to actuate a solenoid. To generate this waveform, we used an ICM7555 Timer IC configured in astable mode. The reference design we relied on [3] shown below indicates the output frequency and duty cycle of the output signal can be configured as follows: solenoid. In parallel with this resistor is an additional MOSFET, which when opened will effectively short across that resistor due to the low drain-source resistance (on the order of 10 s of milliohms) and increase the current through the solenoid to the actuation level. To gate this MOSFET, we use an additional 555 timer, this time in a monostable configuration, which will output a fixed duration pulse, whose width is set by an RC pair across the IC s discharge and threshold pins in accordance with its datasheet [4]. Monostable operation is triggered by a negative/low level signal which must be shorter in duration than the output pulse. To generate this trigger, we use a first order RC differentiator, offset from zero by V CC and with the positive spike generated at the negative edge of the input suppressed by a diode. Below is a simulation model of the circuit, as well as its output: Fig. 6. Differential Trigger Simulation & Output V. MICROCONTROLLER Fig. 5. 555 Timer Reference Design [3] f = 1.44/(C (R 1 + 2R 2 )) D = (R 1 + R 2 )/(R 1 + 2R 2 ) The output of this IC drives a power MOSFET, which gates the power signal driving the actuators. To drop current down to the holding level, we use an additional series resistance between the source and ground of the MOSFET connecting the MCU and The Microcontroller is the heart of this device and is built in the power PCB to control outputs and inputs. There are 2 MCU s to increase speed of processing and divide the tasks of the Bluetooth and tactile sensing functions. A. Microcontroller For this project, the microcontroller will need to have more than fifteen I/O pins as to connect with many components like sensor, Bluetooth, vibration motors and actuators. Clock speed must be more than enough to handle the processing power for the system to 4

function at a fast speed. Flash memory size must be more than the system needed to store the large amount of software so that the device can continue to function normally after restarting. Active power draw needs to be as low as possible depending on the clock speed because this project is a handheld device and all the components need to consume as little battery power as possible so that users can use the device for a long time. The ATSAMD21G18 microcontroller in the Arduino Zero board has a fairly high clock speed of 48MHz resulting a good amount of processing power for this project. With 256KB flash memory, programming can be easier, larger amount of codes can be done. 38 I/O pins will be more than sufficient for all the components to connect to the microcontroller. Low active power consumption of around 10.3mA, the microcontroller will save power when in use resulting longer battery life. Programming IDE is very users friendly, it will help beginners create better projects. Due to the demand of the Bluetooth we decided to incorpate two MCU s into our design as initial tests showed it eased the demand of the initial MCU. a time when they are needed and then after, immediately turn them back off. This will make crosstalk between sensors impossible because only one sensor will ever be on at any given time. The Bluetooth microcontroller is responsible for handling Bluetooth communication to and from the E.A.S.I. smartphone app through the Bluefruit LE UART Friend Bluetooth module. It is also responsible for parsing the navigation data received from the smartphone app and controlling the left and right navigation solenoids and the navigation vibration motor based on the navigation data. The Bluetooth microcontroller will receive data from the primary microcontroller that it will relay through Bluetooth to the E.A.S.I. smartphone app using the Bluefruit Bluetooth module. A. Program Flow Chart B. Function The primary microcontroller receives pulse width values from each of the three ultrasonic sensors and converts the values to inches so that they can be more easily understood. The value received from the distance sensor is used to determine which distance solenoids to raise and lower as well as how intense the distance vibration motor will be pulsed. There are three distance solenoids with each solenoid indicating that the detected object is somewhere within a 3ft range (0-3ft, 3-6ft, 6-9ft). Within these 3ft ranges, the distance vibration motor is used to provide an even more precise 1ft range for the detected objected. For example, in the 0-3ft range, if the object is between 2-3ft, the vibration motor will vibrate with low intensity, in the 1-2ft range it will vibrate with medium intensity, and in the 0-1ft range it will vibrate with high intensity. The values received from the left and right direction sensors are used to determine if the left and right direction solenoids should be raised or lowered. If the value is less than 5ft, then the respective solenoid is raised, otherwise it is lowered. The primary microcontroller will also send data indicating which solenoids are raised and lowered and how intense the vibration motor is vibrating to the Bluetooth microcontroller so that it can be sent to the EASI smartphone app through Bluetooth. To eliminate crosstalk between the three ultrasonic sensors, the primary microcontroller will turn on the sensors one at 5

VI. PCB AND SCHEMATIC DIAGRAMS. Due to the complexity of our project we divided our full design into two PCB s. The first PCB is a combination of the Power subsystem, the MCU s and the Bluetooth circuitry. The second PCB focuses on the outputs of our design and the controllers for our Tactile Feedback System. The PCB schematics and board layout were done in Cadsoft Eagle 8.1. The group quickly found that using Eagle would allow for the quickest production. Libraries for all components used were readily available for most of our manufacturers aside from several components within the Tactile PCB. VI. CASE DESIGN We designed a prototype case to contain our components, and had the design 3D printed. The top, back, and front panels fit in place via dovetail joints, and the solenoids all mount into individual brackets that slide into the top panel. The case is 25 x 7 x 6 cm in volume. The sensors mount to the front panel, the battery fits in the back under the actuators, and the PCB s are stacked on top of one another towards the front. The model for each component was generated using Autodesk Inventor, and below the full set of assembled components can be seen. Fig. 7. Power/MCU/Bluetooth PCB. Above is a figure of the group s finalized PCB with the power circuit, dual MCU s and Bluetooth. Below is the finalized Tactile PCB. The boards were manufactured by OSH Park. The Power PCB was manufactured at price of three for $114 while the tactile PCB was $61 for three. The cost was based on surface area. The Tactile PCB is smaller at 2.3 x 2.6 inches. The Power PCB is the larger with dimensions of 4.5 x 2.5 inches. Components were obtained separately by the group and soldered by the group with Stephen and Hung as our primary Solder Team. Fig. 8. Tactile PCB. Fig. 9. Digital Case Design. VIII. MOBILE APPLICATION. In this section, we will elaborate on our software development process, starting with the Bluetooth integration from the Arduino board, the design and development stages, getting the app perform basic functions, navigations and finally the testing. Since we decided that we were going through with Android as a development choice, it made sense to develop a plan for developing and testing. Android platform is the operating system developed for most smartphones beside Apple and Windows phones. Android is open source and application can be developed by anyone with any computer operating system, not like Apple application that need Apple software and hardware to develop. To develop Android apps, the de-facto languages are Java, C and C++. C and C++ are the two less popular languages to create Android app and often less promoted by Google. Java was developed in response to developers wanting a language that was easier to learn and develop with. With the Java classes, it will be easier to create and maintain the app functionalities. We ended up choosing Java for Android mainly because of massive documentation for the Bluetooth, and all our team members have Android phones. 6

Android Studio is a software to create Android application using Java language, the software is easy to use with millions of different examples and projects for any kind of Android application development, also the software is well documented and will make the creating of the project application easier to start and develop. in our original design and avoided fluff and redundancy in design, and so that users do not have to navigate around the app, just need to turn the app on and using it right away. A. Bluetooth Integration Before any new development was done, we had to make sure that data is being transmitted to the mobile app correctly through Bluetooth. For testing the group used the Adafruit Bluefruit LE UART Friend Breakout that have the Bluefruit LE UART Friend Bluetooth module that will be used in the project and an Arduino Uno. This combination was chosen due to the extensive amount of documentation available. Fig. 11. E.A.S.I. s main screen. Fig. 10. Testing the Bluetooth connection via Adafruit s open source nrf8001 app. This was done before development. B. Design & Development We knew in the early stages of development that E.A.S.I. had to be well designed, centered around its core functionality. A lot of projects typically get lessthan-anticipated results due to being too feature-heavy and not having focus; we wanted to make sure that our core functionality was up to par with what we wanted. As far as User Interface and Experience goes, we wanted an app with a simple, and App Store-ready feel. Implementing with simple and big buttons so that the users can touch the right buttons without seeing where the buttons are. Functionality-wise, we wanted to make sure that the design and development of E.A.S.I. was focused on core features. A user should be able to easily turn on the app and by using the buttons, they can turn on and off some of the features from the device like navigation, audio sound and audio ping. We minimized the number of screens down to one screen C. Basic Functions After defining our scope for development and implementing our overall design, much of our focus was making sure a user could use the app to customize basic functions of the device and receive alerts from it. As the app is turned on, it will automatically connect to the device via Bluetooth connection. The data will be transferred constantly between the app and the device. Scanned object s locations and distances are transmitted to the app always. There will be voice prompts to alert the users of such data, and the Audio Voice button will be used to turn on and off all the prompt alerts from the app. The distance of the scanned objects will also alert to the users with ping pulse sound, the Audio Ping button is to turn on and off those alerts. D. Navigations Navigations will be an extra feature for the users. For the app, we will use the HERE Maps API to integrate the turn by turn navigations to the device. There will be two pre-installed addresses for Home and Work. When the user taps on the Navigation button to turn the function on, there will be voice prompt asking for destination, users then need to speak the destination and with voice recognition, the HERE Map will direct the user by sending left and right turns to the device. There will be tactile feedback on the device to let the user know when to turn left and right to their destinations. Due to the inaccuracy of the phone after extensive testing, navigation was removed as a function 7

E. Testing As far as testing the different features of the app, it was simple since it would be obvious in the display that certain features are working. Since the application is very simple and easy to use, the testing is also easy to check to see if the functions are working as they supposed to. Clicking on the sound and ping button will turn on and off all the sound and ping pulses that alert the users. Pressing the navigation button will have the directions sent to the device and guide the users to their destinations. With all those buttons and functions working properly, we can state that the application is complete and all the data being transferred completely and correctly between the app and the device via Bluetooth connections. IX. CONCLUSION. This project has proven to be an incredibly valuable experience for each person in our group. We have applied the concepts and ideas that were introduced in class to bring our idea into reality. Additionally, the group came to understand all the aspects of engineering that come together to bring a project into fruition. Overall, this experience has shown us about ourselves, our communication skills, our relative skill sets, and engineering. Our device is fully functioning and works well in solving the problem that the group set out to fix. REFERENCES. [1] Battery University, "BU-409: Charging Lithium-ion," 21 02 2017. [Online]. Available: http://batteryuniversity.com/learn/article/charging_lit hium_ion_batteries. [Accessed 07 04 2017]. [2] NXP Semiconductors, "I2C Bus Specification," 04 04 2014. [Online]. Available: http://cache.nxp.com/documents/user_manual/um10 204.pdf. [Accessed 07 04 2017]. [3] ElectronicsTutorials, "555 Oscillator Tutorial," 06 12 2016. [Online]. Available: http://www.electronicstutorials.ws/waveforms/555_oscillator.html. [Accessed 07 04 2017]. THE ENGINEERS. Patrick Shiver is graduating Electrical Engineering in May 2017. He has been working three years at Universal Orlando and over 10 years in the Air Force as an avionic technician. After graduation, he will be working with Northrop Grumman in the field of Aerospace system. His Interests are power systems, integration and test, and PLC control systems. Hung Tran is a senior at the University of Central Florida. He is graduating with a Bachelor s in Computer Engineering May 2017. His interests are embedded software, mobile applications and computer hardware. Heath Cissell is in his senior year at the University of Central Florida set to graduate with a Bachelor's of Science in Computer Engineering with a minor in Secure Computing and Networks in May of 2017. He is particularly interested in the computer and network security fields Stephen Miles was born and raised in Tampa, Florida, and will be graduating with honors in August 2017, with a B.S.E.E and a Minor in Computer Science. His interests lie in the digital design and embedded systems fields, and is looking to pursue a career in these or in FPGA technology and design. [4] Maxim Integrated, "ICM7555/556 General Purpose Timers," 11 1992. [Online]. Available: http://datasheets.maximintegrated.com/en/ds/icm755 5-ICM7556.pdf. [Accessed 07 04 2017]. 8