Walking Robot with Vision (WRV)

Transcription

1 Walking Robot with Vision (WRV) Phillipe Jean-Jumeau, Steven Schultz, Anselet Jacques School of Electrical Engineering and Computer Science, University of Central Florida, Orlando, Florida, Abstract Robots are being used for number of different things in today s technology, therefore many robots with different characteristics and uses can be built depending on their intended function. The objective of this project is to build a hexapod robot that is able to walk, and be to discern certain colors, it is to controlled wirelessly by a PS2 controller, which will dictates the robot s every movement. Index Terms Color-sensor, optoelectronics, RF, robotics, photoresistor, wireless communication. microcontroller will coordinate the robot s movements and direction. The robot will have a DC voltage source as its main power source. II. HARDWARE COMPONENTS The robot mainly consists of 4 subsystems. They will be discussed with more details in this section. They are: The robot Playstation 2 Controller Interpreter Wireless Communication SSC-32 Servo-Controller Video Processing Unit Power Supply I. INTRODUCTION In today s society robots are everywhere, whether is it just simply a toy design to entertain children, a computer that performs a task at a scheduled time, a machine that is supposed to help workers perform his or her job faster and better, a car that can take someone from point A to point B without the use of a driver. One good example would be a taxi cab that can drive you anywhere you want within a certain area. Taking a page from cinema, a robot can be designed to be a spy to be used for a number of operations. All these different types of robots have one important thing in common; which is a certain level of intelligence; a program that tells it what to do or look for, and when to do it. In a few decades from now, robots will probably be as commonplace as seeing someone, being led by a dog on the streets. The robot we are building is intended to be a coloredobject detecting robot, with some of its functions being controlled thru a PS2 controller. It is shaped similarly to a spider, with six legs and a specific-color object sensor. The purpose of the robot is to be able to distinguish between certain colors of objects of which it will walk over to detect them. In order to do that, the robot has a color unit that is composed of an array of photo-optic devices with colored-light filters. The color-sensing object is housed on the underside of the robot body. While it is looking for a certain color which is going to be its target; once the targeted object is found, a corresponding red, green, or blue LED will turn on to indicate target acquisition. A Fig. 1. A. The Robot Generic block diagram of the whole system. The robot s main body is a circular chassis colored in black bought from Lynxmotion, it is capable of have six legs attached to it, it contained plenty of room to store necessary electronics and met necessary weight and size specifications. In fact, this chassis was a bit smaller, and weighed a bit less than the other chassis looked at for the body of the robot, and also it enables the robot to have

2 great degree of mobility, because the robot is able to move in any direction at any point of in time sans rotation since the chassis is circular. space. The power supply of the robot meets the power requirement of every subsystem, and allow for future project expansion if needed. Fig. 4. An example of the robot s servomotor. Fig. 2. Robot s Chasis. The robot has six legs, each leg has three degrees of freedom: hip, knee, and femur and each of them has three servo motors for a grand total of eighteen servo motors. The servo motor used for the legs is the HS 475HB Deluxe from Hitec; each servo motor includes Karbanite gears, one piece printed circuit board, and a ball bearing. The dimensions of the legs are: Hip Hor. To Hip Vert.:29 mm, Hip Vert. to Knee Vert. (femur):57 mm, Knee to Foot (Tibia): 141 mm. The servo motors are controlled by an SSC-32 servo controller; it uses an ATmega168 microcontroller. The SSC-32 servo controller provides the walking algorithm for the robot. B. Playstation 2 Controller Interpreter A Playstation 2 controller will be used to control wirelessly the many functions of the six-legged robot, including its motion and its decisions concerning targeting. To one s comfort, the Playstation 2 controller provides two analog joysticks, a control pad (the cross looking buttons), and twelve buttons, including the ones that are pressed by pushing down on the joysticks. The Playstation 2 Controller Interpreter is the circuitry that permits the communication between the electronics on the robot and the Playstation 2 controller. The brain of the Interpreter is the Atom Pro 28, a microcontroller by BasicMicro. It is preferred over other microcontrollers by many robotics enthusiasts, due to the fact that it is equipped with 4KB of EEPROM, which is ample enough to contain formulas, variables and constants for the necessary calculations, and the fact that it can be programmed using BASIC language and has a simple architecture, hence making it a simpler microcontroller than most of Atmel s and Microchip s microcontrollers. Fig. 3. An example of the robot s leg. The robot s power supply is fairly efficient and allows the two main battery sources to run as long as possible - about 30 min of continuous use - before needing recharge. The power supply also takes into consideration cost, time, and Fig. 5. BasicMicro s AtomPro 28 Microcontroller. The Playstation 2 controller will be communicating with the Atom Pro microcontroller through 4 I/O pins. They

3 will be connected to the controller s DATA, COMMAND, CLOCK and ATTENTION pins. The DATA and COMMAND lines carry byte frame data; the first is a oneway traffic from the controller to the microcontroller, whereas the second from the microcontroller to the controller. The DATA line transfers data associated to the buttons pressed, while the COMMAND line permits the Atom Pro 28 to configure the PS2 controller. In this case, it will set the controller in Analog mode, which will allow the use of the joysticks. The CLOCK and ATTENTION lines, both being controlled by the Atom Pro microcontroller, synchronize the data in the DATA and COMMAND lines (transmitting and receiving of data). The microcontroller processes the data received from the DATA line, then outputs the commands corresponding to the button pressed to the unit on the robot, which, in this case, will be transmitted wirelessly. The data transfer rate between the Interpreter and the unit on the robot will be set at bps. C. Wireless Communication The robot will be controlled wirelessly by the Playstation 2 controller. Since the data traffic is one-way (from the Interpreter to the robot), the wireless system only consists of one transmitter and a receiver. The transmitter is implemented by using the TXM-916- ES RF transmitter module and the ANT-916-CHP Chip Antenna from Linx Technologies; both of them operate at a center frequency of 916MHz. The TXM-916-ES RF chip features FM/FSK modulation, the latter being necessary to handle the data from the Atom Pro microcontroller. It can also handle data rates up to bps, which is quite convenient since the servo-controller and the controller interpreter will be communicating with each other at a data rate of bps. Fig. 7. The ANT-916-CHP chip antenna. The SSC-32 (serial servo controller) controls the servomotors and provides them with a range of about 180. It provides 32 channels for servomotor control. It is built around an Atmega 128 microcontroller and an 24LC32P EEPROM which contains the firmware. A DB9 input is provided for communication with a PC which is used to program and configure the servocontroller and the position of the servomotors for our purposes through a software called PowerPod, provided for free by Lynxmotion. Fig. 8. The SSC-32 Servocontroller. It is powered by two different power sources: One for the electronics and ICs on the board, the other for the servomotors. The electronics and the ICs can be powered by a regular 6 or 9V battery; however, the servomotors need to be powered by a 6V battery with a higher power output. Fig. 6. The TXM-916-ES RF transmitter module and the RXM-916-ES receiver module. The same antenna and the RXM-916-ES RF receiver module, also from Linx Technologies, are used for the receiver. Like the transmitter, the receiver uses FM/FSK modulation and is capable to handle data transfer rates up to bps. III. POWER SUPPLY Discussions and research provided insight into the need for appropriate component selections and placement, ensuring a proper and successful implementation of a DC power converter, and printed circuit board implementation. The power supply design for the robot takes these specifications into account the following:

4 I. The main voltage source of the robot are two DC Batteries II. The source voltage is step down using voltage regulators III. The power supply system accounts for noise issues and interferences IV. Each subsystems have different voltage distortions V. Isolation is used to reduce the effects of noise and interferences between the subsystems The following materials will be needed as well: voltage Regulators, 1 on/off Switch, capacitors, 2 batteries, resistors, 5 LEDs. The maximum power requirement of each subsystem is classified in table 1 below and so is the overall maximum power requirement of the robot. The table helped us decide the battery source voltage and the type of regulator that is needed to reduce the source voltage to the required voltage for each of the subsystems. To decide the voltage of the battery source, we also include the losses in the lines, and losses cause by heat and interferences. Between a 5% to 10% loss factor to our battery voltage source to compensate for losses when the robot is operating is included. Since the highest voltage of the different subsystem is 9V, it was decided that the battery source would be around 12V. TABLE 1 SUBSYSTEMS REQUIRING VOLTAGE, CURRENT AND POWER Control Unit Voltage (V) Current (ma) Maximum Power (W) Micro Controller Micro- Controller for color sensor Micro Controller TAOS TCS230D color sensor Micro Controller Receivers Transmitter Servo-Motors (18) From the different subsystems of Table 1 above, the power requirement for the robot is broken down into two main subsystems, the servo motors subsystem and the electronics subsystem. Therefore the robot uses two rechargeable batteries as its main voltage source, one to power the servomotor subsystem and the other to power the electronics subsystem. The power supply consists of a 6V 2800mAh rechargeable Nickel Metal Hydride battery to power the robot s eighteen servo motors, and a 12V 1600mAh rechargeable Nickel Metal Hydride battery to power the electronics. The voltages are being regulated by four LM317T voltage regulators, eight capacitors, four. 1mF and four 10mF to provide stable constant and continuous voltages, and LED that serves as a power indicator and an ON/OFF switch to to turn the robot on and off. Table 1 above gives the required power for the different sub-systems of the robot. A. Purpose IV. COLORED-OBJECT DETECTION As stated previously, an objective we set forth to accomplish was to design and implement a system to discern objects of specific colors. The colors we are trying to detect are: red, green, and blue. The assumption was, in choosing these three colors, was that they are primary colors they should be the easiest to detect. Being able to detect objects of different color has a myriad of applications. The most obvious would be for sorting purposes for robotic machinery. For instance, being able to decide if a targeted object is desired and then allowing the user to choose a course of action (or performing a pre-programmed plan of action) is a powerful tool. We have also gained a good understanding of how sensors work and how they can be used to model real-life systems. B. Background and Theory An obvious solution to our objective would be a video camera using either real-time video processing or a framegrabber with image processing. A small CMOS camera (typically used in cell phones) interfaced with a Field- Programmable Gate Array (FPGA) was considered at first for the speed of operation and efficiency. However, the difficulty of the programming language VHDL and the time constraint of this project made the camera and FPGA option unrealistic. For image processing, a microcontroller was also considered. However, most microcontrollers have too

5 small of memory to hold video frames coming at 30 frames per second. The design would have to discard frames or lower the image resolution of the camera artificially by skipping rows and columns of pixels. Also, two microcontrollers would have had to been programmed: one as a framegrabber and one for the image processing/color blob detection algorithm. This was also unrealistic for the time allotted for the project. Thus, we decided to go with a simpler option using optoelectronic sensors/devices. Optoelectronic devices are devices that respond to and function according to different amounts of electromagnetic light detected. Whether it be optical-toelectrical or electrical-to-optical, optoelectronic devices serve as valuable sensor in robotics and many everyday consumer products. For our purposes, the optoelectronic devices in question are photoresistors and phototransistors. Photoresistors are devices that act similarly to resistors but when light in the visible range strikes its surface its resistance level decreases. The amount of dark resistance and light resistance are the parameters involved instead of just a single resistance value. The resistance value decreases (ideally) linearly with an increase in light intensity. Fig. 9. An example of a photoresistor. The curved trace is the optically-sensitive element of the device. Phototransistors, likewise, work very similar to their non-optical counterpart the bipolar junction transistor or field-effect transistor [2]. Phototransistors have only two terminals, a collector and emitter. A transparent cover allows light to enter the device and raise the gate voltage by photon-electron interactions. Thus, when enough photons enter the device the gate voltage will raise and current will be allowed to flow between the collector and emitter. Fig. 10. A phototransistor. Note only two terminals are present (collector and emitter); the base is housed in the transparent head to allow photons to enter and raise the gate voltage. Using either of these devices and colored-light filters, we can build a light sensitive circuit that can sense a specific color and react accordingly. Placing red, green, and blue light filters over our optoelectronic devices we can create a red light sensor, a green light sensor and a blue light sensor. These colored-light sensors act as a bandpass filter for visible light. Red-light filters allow red light to pass over a narrow wavelength range (plus or minus some center wavelength), and then same for blue-light and green-light filters, respectively. To get the correct colored light to the sensor, LED light is shined on to the object to be sensed. A white LED could be used for this purpose (since it contains all wavelengths of the spectrum) or an LED of the same color as the object to be sensed. The light reflected off the object will be the same color as the object; all other wavelengths will be absorbed by the object. Thus, the red, blue or green light will be accepted into the appropriate sensor which will be turned ON by the user. If the reflected light is able to pass through the ON sensor, an indicator LED will turn on, signaling that the correct colored object was detected. This application of the physical characteristics of electromagnetic waves is what allows us to artificially create eyesight for our robot. C. Implementation Options The objects to be detected must be nearly-flat so the robot can walk over them safely as well as be red, green or blue in color. Deciding which color is to be detected is done by the user via switches on the top of robot s body. For red object detection the user will turn the RED switch to ON and the BLUE and GREEN switches to OFF. Similar operation is done for the blue and green object detection. As mentioned previously, implementing vision with a camera and FPGA or microcontroller proved to be overlydaunting because of the time constraint. The remaining options were to use the optoelectronic devices and even under this umbrella there were more options. The first being a circuit consisting of photoresistors with colored-

6 light filters. Reflected light from the object would be incident on all the sensors (photoresistor and filter combination) but only the sensor with the compatible filter would be allowed to trigger a detection response (i.e, the indicator LED). The reflected light incident on the corresponding filter and photoresistor pair lowers the resistance of the photoresistor allowing a high voltage to be placed on the output (the indicator LED). Logic gates could also be used to ensure that the other two filters do not light up the indicator LED (in the case of an inbetween color). Logical AND gates and NOR gates could be used to ensure this. Another implementation would be using the phototransistor. Red, green, and blue colored-light filters would also be placed on the phototransistors similarly to the photoresistors. Once the color filter and phototransistor combination receives light the phototransistor will allow current to pass through a resistor to create enough voltage to turn on the indicator LED. In both of the above implementations electrical tape was placed around the photoresistor and phototransistor to block ambient light from entering and causing false readings. Also, as the light source of the reflected light, the white LEDs proved to be too intense and almost always gave a detection regardless of the colored object. So, red, green and blue LEDs were used depending on what the user selected. The intensities of the different colored LEDs were equalized with different resistor values. When tested, both of these implementations proved problematic. Photoresistors and phototransistors alike gave too varying of values from one to the other as the consistency from individual device-to-device was low. The object was either too illuminated or ambient light was too much of a factor. With the wide-ranging parts/devices available, the circuits became too erratic. A third and best solution was found: the TAOS Inc. color sensor, the TCS230D. The TCS230D converts light-to-frequency. Increasing intensity of the selected color gives increased frequency of square wave pulses generated. Thus, using a simple microcontroller we can count the number of square waves generated during a given sample time and determine the color being sensed. The TAOS color sensor incorporates an 8 x 8 array of photodiodes as can be seen in Figure. Three groups of sixteen of them each have red filters, blue filters, and green filters. The fourth group of sixteen have no filters. S0 1 S1 2 OE 3 GND 4 8 S3 7 S2 6 OUT 5 V DD Fig. 11. The top view schematic of the TAOS TCS230D Color Sensor. This schematic shows the pins and an inside view of the photodiode array. The pins S0-S3 are used to select which color is to be detected. This turns on the corresponding set of photodiodes. Like the simple circuit implementations before, the TAOS color sensor still works on the principle of using reflected light off of the object, but the TCS230D is much more accurate since the color filters are of better quality and made for the device. The distance the TCS230D must be from the object is also about the same as the simpler circuits: about 2-3 inches. More than one indicator LED will have to be used now since the TCS230D has only one output. The select lines are connected in parallel with three indicator LEDs each corresponding to the correct color. The other two will be shut off from receiving any voltage so as to not give a false indication. The final, missing piece of our color sensing system is the microcontroller to count the number of pulses generated by the TCS230D. Something small in power footprint and easy in programmability would be appreciated since the task at hand is relatively simple: count the number of pulses outputted and signal the corresponding indicator LED. A microcontroller programmable in C or BASIC and did not require much setup would be ideal. The ATmega644 from Atmel and the PICAXE18X from PICAXE (a variant of the PIC from Microchip) were compared. The ATmega644 had the obvious advantage of more memory and faster speed but for our purposes provided too much horsepower. The PICAXE18X also featured a count feature that can count the number of pulses received in a given sample time. For programming the PICAXE18X, BASIC language was used and the sampe times were defined. Since the TCS230D uses photodiodes, the stabilization time was very quick so a short sample time was needed. Speed was never a specification though because the robot is not a very fast one.

7 TABLE 2 COMPARISON OF MICROCONTROLLERS ATMega644 PICAXE18X Clock Speed 20 MHz 4 MHz Memory Size Power Consumption 64 kb of program memory, 4 kb SRAM, 2 kb EEPROM 240 micro- 1 MHz (3 V) About 1000 lines of program code. 150 micro- 1 MHz (2-4.5 V) The code to program the PICAXE18X starts up the device, has a delay and then actively looks for pulses during the sample times. When the robot is walking and no object is underneath it, the color sensor will not provide a strong enough output to the microcontroller. The microcontroller will compare it to the accepted threshold and discard it if not with in bounds. V. CONCLUSION The Walking Robot with Vision is an example of how different disciplines of engineering can come together to work in unison. Taking advantage of physics the properties of natural phenemona also play a part in how simple (or complicated) a design can become. We hope to continue tweaking our robot and offering a more advanced vision system. In the future, we expect more and more projects such as these to meld engineering disciplines together; blurring lines. ACKNOWLEDGEMENT We would like to extend our greatest appreciation to our family, friends and loved ones for supporting us through the length of the project. And of course, our deep appreciation to Dr. Sam Richie and the faculty for their guidance. REFERENCES [1] Interfacing a PS2 (PlayStation 2) Controller. Curious Inventor < ps2> [2] Held. G, Introduction to Light Emitting Diode Technology and Applications, CRC Press, (Worldwide, 2008). Ch. 5 p 116. ISBN: BIOGRAPHY Anselet Jacques, a senior student of the Electrical Engineering Departement at the University of Central Florida, will start to persue his masters degree at the University of Central Florida after graduated in the Fall of He also takes part in the intramural sports ( soccer) activities at the University of Central Florida. Steven Schultz is a senior majoring in Electrical Engineering with minors in Mathematics and Physics at the University of Central Florida. His academic interests include digital signal processing, biomedical engineering applications, and audio/musical signal generation and processing. He enjoys film, playing guitar and reading in his spare time. Steven is currently working as a student co-op for L-3 Communications Link Simulation and Training. Phillipe Jean-Jumeau is a 21-year old Electrical Engineering student at the University of Central Florida. He will pursue his Masters and doctorate degree in the same field.