Carl Hayden High School s Falcon Robotics Autonomous Underwater Vehicle Entry For The RoboSub Competition Sovereign Rover

Transcription

1 Falcon Robotics AUV Carl Hayden High School s Falcon Robotics Autonomous Underwater Vehicle Entry For The RoboSub Competition 2011 Sovereign Rover Abstract: Falcon Robotics Sovereign Rover is the team s first effort in the autonomous underwater robotics arena. The team has been involved in teleoperated underwater robotics for seven years, having competed in the Marine Advanced Technology Education s National ROV Championships for three of those years. In addition, the team has been involved for ten years with FIRST Robotics (For Inspiration and Recognition of Science and Technology), a land based teleoperated robotics platform with limited autonomous capabilities. The team has achieved success despite facing problems typically associated with schools struggling in an inner city environment. Bringing a great deal of robotics experience to bear on the challenge presented by the RoboSub Competition, the team hopes to use a systems integration approach to solving problems. The goal is to avoid inventing components, but, instead, try to rely on off the shelf products whenever possible. After observing and researching the RoboSub Competition over the past three years, as well as reviewing the journal papers, we feel prepared to take on the challenge. We believe that by using a Windows 7 computer with various sensors such as forward and downward facing cameras, depth sensors, an all-in-one digital compass, gyroscope and three axis accelerometer, that we can complete many of the mission tasks. We also want to demonstrate that building a capable AUV can be relatively inexpensive. Authors: John Rangel, Eduardo Fernandez, Maria Castro, Bianca Rodriguez, Isabel Martinez, Dillon Dayea, Diserae Sanders, Carlos Ruvio, Martin Carranza, Jesus Meraz, Jorge Tay, Elizabeth Perez, Daisy Fernandez, John Harris. 1

2 Introduction: Sovereign Rover was designed to compete in the 2011 Association for Unmanned Vehicle Systems International and Office of Naval Research Autonomous Underwater Vehicle Competition. Since this is our first year participating in this competition, we designed our robot to be as simple and efficient as possible. We have been participating in teleoperated underwater robotics for seven years now. By relying on our experience in underwater robotics, we were able to build our AUV with time tested features we had previously developed. This allowed us to focus on our weakest point - programming. Our vehicle frame is comprised of an octagonal shape made of pultruded fiberglass with motors mounted at angles so we can use a vector drive system. We feel that this configuration gives us the most maneuverability. Because we wanted to keep our robot as simple and inexpensive as possible all of our motors are off the shelf bilge pump motors. Our control system is also unique in its simplicity. We are controlling the majority of the functions for our AUV using Phidget components. Phidgets are a set of plug and play building blocks for low cost USB sensing and control via a personal computer. The best part is that Phidgets eliminate the need for interfacing devices and additional electronic circuits. This allows our electrical components to be easily controlled by code and keeps the overall AUV electronics simple. The vision system, however, does not use a USB interface. Instead we chose to go with Allied Vision Technologies machine cameras with 1394b (fire-wire) because of the superior performance over webcams (Fig.2). Mechanical: The frame for Sovereign Rover consists of an octagon made of ⅛ in. thick 90 degree angle pultruded fiberglass produced by Creative Pultrusions. We used fiberglass because of its durability and because it is non-corrosive, non-magnetic, with high strength to weight ratio. To make sure the sub is upright at all times we have the electronics housing, which is the most buoyant compartment, located on the sub's highest position possible (Fig.1). The frame also accommodates two cameras, forward and down facing. They are mounted on the frame via a piece of 4 in. wide, ⅛ in. thick c-channel fiberglass. Slots are cut in the c-channel where the camera housing fits. These slots prevent any movement of the camera housing and 2

3 zip ties lock the cameras into position. The torpedo firing mechanisms are located on both sides of the front and down facing cameras allowing for the most accurate aiming possible (Fig. 3). The reason for the octagon shape is to allow the thrusters to be mounted at forty five degree angles to the square geometry of the frame (Fig.1). We call this thruster configuration Vector Drive. This will allow Sovereign Rover to maneuver in any direction, depending on the combination of thrusters that are powered. The AUV was designed so that the water flow of the thrusters is minimally impeded by the AUV frame or components. This allows us to get the most out of the thrusters. Harness: Sovereign Rover uses twelve bilge pump motors (Rule 1100GPH/4160LPH 12V) (Fig.4). Four are used for vertical movement and eight for horizontal movement. These motors are small and light weight and provide 2.1lbs of thrust, which is 48% of the thrust of one SeaBotix thruster, an industry standard for competition AUVs. Knowing this, the team put two thrusters on every thruster position, making the thrust almost equal to one SeaBotix thruster. Designing the sub with these smaller and economically sound thrusters costing only $35 each instead of using SeaBotix thrusters saved the team $4,800. We devised a harness to help with the deployment of the AUV. We mounted four stainless steel eye hooks on the corners. The nylon harness consists of four stainless steel carabiners attached to each of the four ends of the rope. All the ropes were then secured to a steel ring creating a four point harness with one lift point (Fig. 5). Electronics Housing: Learning through collaboration with other robotics teams in the area, we have incorporated water proofing methods which are new to our team but have been used and proven to work by other teams. Our battery and camera housing are examples of this. We learned this method from a team in 3

4 Sierra Vista, Arizona, known as The Typewriter Repairmen. The concept is easy: A transparent polycarbonate tube with machined end caps. The end caps have an o-ring groove machined in them so that the o-ring is pressed against the inside surface of the polycarbonate tube. Having a clear tube allows us to visually verify that it is properly sealed. Both the camera and battery housings are made in this fashion, but the means of connecting them to the main electronics housing differs. For the battery housing, we simply used SubConn connectors. Choosing a connector for the camera housing was more difficult. To our knowledge, commercial water proof fire wire cables are not available. Cutting into the fire-wire cable would have compromised the quality of the signal of the cable. The solution was to fuse a small ½ in. diameter brass pipe on each end of the cable using 3M Scotchcast (Fig.6). space for an o-ring to set in and make a seal. To ensure that the o-ring stays put, an aluminum plate with a ½ in. hole sits over the brass pipe to secure the o-ring. The main electronics computer housing was originally intended to be made of aluminum, but due to time constraints and manufacturing difficulties, the team decided to make it out of acrylic. Acrylic is half the weight and has desirable properties such as being non corrosive, transparent and easily machined. The housing consists of three parts: the top lid, the bottom lid and the main compartment (Fig.7). The main compartment is rectangular tube with 2 in. lips on two sides. On these lips an o-ring groove was machined. This allowed the o-ring to be sealed against the acrylic lids when bolted shut. This method makes a great seal that can be seen through the clear acrylic. The brass pipe gave us a smooth cylindrical surface for an o-ring seal and the Scotchcast sealed the cable to the inside of the brass making it waterproof. We then drilled a ½ in. hole on a flat surface of the bulkhead of both camera and computer housings. Then we drilled a 3/5 in. hole right on the top of the existing ½ in. hole, giving us the perfect Torpedoes and Darts: Because we are trying to use off the shelf products as much as possible we decided to use Toypedo Blast from SwimWays for our torpedo mechanism as well as the dart dropping system. The toy gun is basically a cylinder in which you 4

5 load a torpedo and a plastic latch preventing it from firing. We use a servo to pull up the latch enabling the torpedo to shoot. The torpedo itself is a rubber projectile that is shaped like a rocket. Spare parts can be easily obtained since it is an off the shelf product. We disassembled the Toypedo Gun in order mount it to the AUV on each side of each camera to take up less space. We attach a very thin piece of foil that can be ripped easily when the gun is fired so as to stop the darts from falling out at the wrong time (Fig.8). Buoyancy : In our vehicle we distributed the mass and the buoyancy so that the AUV maintains a stable and upright position at all times. The vehicle is.5 percent positive in its buoyancy as required by the rules. Due to differences in bodies of water, extra mass will be added if the AUV is too positive or extra buoyancy added if the AUV is too negative. The bulk of the buoyancy is from the main electronics housing and the majority of the mass is the rest of the AUV. The fiberglass is slightly neutral and doesn't play a big part in the buoyancy or the mass. Magnetically Controlled Switches: There are three switches outside the AUV that control functions. First is the power switch, next is the mission start switch and third is the computer on/off switch. We use magnetic switches because we want to maintain the water tightness of the case. For the power switch we are using a low voltage circuit to control a relay that switches the high voltage circuit. This switch is also called the kill switch because the diver can remove the magnet to shut down the power from the batteries (Fig.9). In order for the AUV to start the mission, the mission start switch activates a logic circuit that either sees a true or false condition on our Phidget Interface Board. To turn on or off the computer, the third magnetic switch is used to close a circuit acting as a standard power button on the mother board. These three switches allow us to eliminate the possibilities of leakage in our AUV while not reducing the desired functions. Phidget Control system: To control the twelve thrusters in the Sovereign Rover we used six motor speed controllers from Phidget (Model #: 1064-PhidgetMotorControl HC). Each motor speed controller is capable of controlling up to two high current DC motors. They control the AUV s angular 5

6 velocity and angular acceleration. The thrusters are connected to the motor speed controller board through the electronics case with SubConn connectors and the motor speed controller is connected to the main computer through a USB cable via the Phidget interface kit. Each Phidget motor speed controller has a 15 amp auto reset fuse in-line to protect them. (Fig.10) Phidget Servo control: The four servos that actuate the releasing mechanisms for the torpedoes and darts on the AUV are controlled through the 1601 Phidget Advanced Servo 8 controller. It has a built in switching power supply that allows the unit to operate from 6 to 15 VDC. The servo controller allows the control of the position, velocity and acceleration of the servos connected to it (Fig.11). Servos: The servos are each enclosed in an ABS pipe with a SubConn connector. Each servo is connected to a ⅛ in. diameter brass rod with an aluminum coupler. The brass shaft sticks out through the ABS enclosure through a ⅛ in. shaft seal from Mike s Subworks, a company that specializes in model submarines. The shaft seal is affixed to the ABS enclosure with JB weld (Fig.12). Phidget Sensor input: We have six Phidget sensors in our AUV, a light sensor, a current meter, a volt meter, and a three in one digital compass, accelerometer and gyroscope. All these devices come with sample code libraries that could be integrated in to our AUV code. The light sensor measures human perceptible light in lux. It has a range from 1 to 1000 lux. The current sensor, can measure up to 50 amps DC and is ratiometric. Ratiometric refers to the ratio of the output voltage to the supply voltage. The voltage sensor measures the proportional differential between the input and output terminals. The Spatial sensor has three functions: a three axis compass, a three axis accelerometer, and a three axis gyroscope (Fig.13). The sensor is plugand-play hardware. We integrated it into our system using the code libraries supplied 6

7 by Phidget. At start of the mission the compass auto-calibrates once the mission switch is activated. We use it to establish and maintain headings. The compass is vulnerable to electro magnetic force (EMF) interference which is problematic because our thrusters produce a significant amount of EMF. Our solution is simple: we only get readings from the compass when the AUV is at rest, when the thrusters are off and little or no EMF is being produced. To reduce drift, we use our accelerometer feature to check velocity. Most of the Phidget devices go to a Phidget Interface kit (Fig.14), that acts as a USB Hub, analog input and digital input and output. Depth Sensor : The depth sensor we implemented is the US PG from Measurements Specialties. We decided to go with this model because of its low cost and compact size. The sensor can operate in temperatures ranging from -40 degrees to 105 degrees Celsius and can measure up to 500 psi (Fig.15). For our sensor to be functional it needs to be connected to a Phidget 4-20mA Adapter which produces 15 volts to power the sensor. The adapter is then connected to a Phidget Interface kit through an analog interface. The adapter measures the current passing through the attached depth sensor and converts it to a voltage, where it is then displayed as a sensor value. This is then calibrated to a known depth. Machine Cameras : We are using two Guppy F-033B/C Fire wire Allied Vision Machine Cameras. (Fig.2) We chose to use them because they are relatively inexpensive and very compact without reducing the quality of the image. They were also recommended to us by the Cornell University team. The cameras were made for robotics applications, they are color and have a 696 x 494 resolution at 58 frames per second frame rate. Software: We chose to go with Visual Studio C# language and environment because it is beneficial to development and testing. Using a combination of buttons, sliders, and menus, we are able to successfully test our code faster than we normally could in other languages. Another benefit of using Visual Studio is its helpful auto correct suggestions and automatic imports of code libraries it makes when writing code. Another advantage to using C# is its popularity. We wanted to make sure that the language we wrote our code in was very well known so that if we were struggling on a specific part of the code, we could ask for help. Popularity of the language also meant 7

8 that most libraries were compatible with C#. Thanks to the help of Visual Studio C#, the code was relatively easy to develop and test. Control: The control system of our robot consists of data from vision, compass, 3axis accelerometer, depth sensor, light sensor, and PID motor control. Using a combination of all these features, we are able to accurately control our robot. The compass and depth sensor help us keep our heading and position while the accelerometer lets us automatically adjust if we start to drift. Most of the movement, however, is based on vision. For example, if the AUV sees a buoy, it will automatically adjust to center itself. Thanks to PID control we developed, we are able to have it center itself accurately and efficiently. Vision Processing: Vision processing is the most challenging problem that teams must solve in this competition. Due to the high level of difficulty in vision processing and object recognition, the majority of teams use a library of code made by other people or organizations such as Open CV. Since our code is in C#, we chose to use AForge.net. AForge is a well developed vision processing library for C# with many of the vision algorithms needed to accomplish all the mission tasks. Shape detection is a large part of the vision processing needed to complete the mission. The shape detection algorithms and examples provided by Aforge allowed us to accomplish this relatively easily (Fig.16). Due to variable lighting conditions, we have added a special feature to our color recognition software. We call this feature Filter Scanning. We created a series of color filters that apply to the various times of the day and lighting conditions and save those values. When we want to detect red, for example, the robot will cycle through the color filters of the saved values until the object it is looking for is detected. If it loses the object after a certain time, it will start cycling through the filters again until the proper filter is found. This special feature allows the robot to adjust to various lighting conditions during the mission. Batteries and Battery Charging: The AUV is powered by four 6.5 amp hour 11.1 volt lithium polymer batteries which were chosen due to their power density and high efficiency under stress and strain (Fig.16a). The batteries are separated into four battery compartments. The four compartments are connected in parallel so that the overall 8

9 voltage remains the same but the amp/hour capacity is increased. We effectively have a 26 amp hour battery at 11.1 volts. This will allow us to have a longer lasting battery pack (Fig. 16b). The charger is capable of producing 10 amps while charging. The charger comes with a battery balancer which maintains the battery cells within normal operating parameters, keeping the battery functional. Lithium polymer batteries are known for their particular charging requirements, so care must be taken while recharging them. A critical issue with lithium polymers is swelling, which occurs when gas is generated during charging or discharging when the battery is faulty. The battery balancer is supposed to monitor this faultiness and help prevent it. Despite having this balancer, batteries should never be left unattended while charging. If swelling occurs, salt water can be used to neutralize the battery. Electrical system: As previously stated, we did not attempt to make all of our components. Instead, we used off the shelf components and figured out how to piece them together to make a fully functional AUV. (Fig.17) All the AUV systems are powered from two 12 volt busses, a positive and a negative buss. The computer is run from an inverter that is connected to the 12v buss and then plugs into the computer power supply which powers the computer. Most of our components are mounted and wired inside water tight electrical housing; however, some components, such as the cameras, motors, and the three axis accelerometer, are mounted outside and away from the housing. We used watertight SubConn connectors to connect these components. To ensure that spliced ends of cables are not in any danger of leaking, we covered all exposed wire with epoxy resin, followed by Scotchcast sealant. Taking these precautions ensures that our AUV will be leak proof and fully functional. Computer: The on-board computer we are using is the EC7-1817LNAR. It features a Mini-ITX SBC with Intel I7 Sandy Bridge 64bit processor operating at 3.0 GHz for very high performance (Fig. 18). This combined with 2 DDR3 memory slots with up to 16GB of memory and internally integrated graphics controller ensures that we have few limitations in terms of 9

10 computing power. Utilizing up to 12 USB ports is a great benefit because we have many devices that have USB interfaces. This allows us to minimize the use of hubs. Overall, this very compact and powerful computer allows us to successfully operate our complex AUV. Mission Planner: When our mission commences, our AUV has many tasks it must accomplish in specific sequences. Therefore, we have a mission planner integrated into our code. Our mission planner is unique in that it is not really a decision maker but a task scheduler. We have step by step tasks in the mission planner that lets the AUV know what task it should currently be doing. It also ensures that if the AUV can't complete a task within a prescribed time limit, it will perform a recovery sequence or move on to the next task. Donors, SubConn, Intel Corp., Phidgets Inc., Canyon State Plastics, National Mall Fronts, Xtreme FX,!nventivity, SEAD, Creative Pultrusions, Microchip, Southwest Fasteners, Sims Metal Management, Boeing, and Fastsigns have all been contributors. Most of all we would like to thank JB Rajski of Cornell AUV Team. His valuable assistance throughout the past year has helped bring us into the world of autonomous underwater robots on very sound footing. Front View Angle View Acknowledgments: The would like to thank the many sponsors that help make it possible for us to take part in programs like this. Arizona Tax Credit 10