1
Team RoboBoat 2015 Abigail Butka Daytona Beach Homeschoolers Palm Coast Florida USA butkaabby872@gmail.com Nick Serle Daytona Beach Homeschoolers Flagler Beach, Florida USA Abstract This document describes the S.S. Minnow entry to the RoboBoat competition. Topics in this document include: the team history, experience from the RoboSub competition, overall strategy to accomplish the competition tasks, vision processing used for multiple tasks, and the hardware chosen for the. The report concludes with technical concerns and competition tasks the team feels they can achieve. Keywords Autonomous, RoboBoat, Sonar, PixHawk, APM Rover I. INTRODUCTION To begin, let s review the team history. We first began as a SeaPerch team[1] in 2013. We won 3 rd place overall at the regional competition. That allowed us to go to the National competition in Indianapolis. While at the National competition we won 3 rd place middle school class for the Deep Water Challenge. That year we also decided that we would enter into a new competition, RoboSub[2], which is an international college level submarine competition. During our first year in RoboSub we won 11 th place and the Bang for the Buck award[3]. In 2014 we went back to SeaPerch again. That year we won 1 st place overall for middle school class in the Regional competition. At the national competition in Hattiesburg, Mississippi. There we won 3 rd place overall for middle school class. In 2014 we decided to go back to RoboSub again. We went through the qualification gate, bumped a buoy (this required matching our depth and position to the buoys), and circumnavigated the second gate. Out of the 39 teams at the competition only three teams circumnavigated the gate, the other two teams wound up in the finals. We won 11 th place[4]. In 2015 we were attracted to RoboBoat because of the Interoperability challenge. We wanted the experience of working with quadrotors and learning how to use GPS for navigation. What we learned from the SeaPerch and RoboSub competitions will help us with the RoboBoat competition A. Challenges we faced in RoboSub We struggled with many things during Robosub. Leaks in the box was one of them. The leaks were caused by not checking the rubber seals on our waterproof boxes. Since we are underwater we have no GPS so we are more reliant on dead reckoning and vision. The problem with vision is that it is quite hard to have a vision algorithm that works for all different scenarios. Depending on the time of day and if there is fog changes the way you find objects in the water. The water at the TRANSDEC, the sonar testing pool that is used for the Robosub competition, is very murky and has a lot of algae floating in the water. All of those things contribute to making vision harder. We can t get feedback on what the sub is doing and we also don t know where it is in the water. Looking through the water is hard to do. When we balanced our vehicle we had to balance it on all three dimensions. Same for when we are trying to control our vehicle, we can control it in all three dimensions including depth B. What we learned from RoboSub A few things we learned from Robosub is that when working with waterproof boxes, remember to check the seals of the boxes and make sure that there aren t any wires in them. Another thing we learned is that we should test algorithms and new sensors extensively to make sure they can do what we need them to do. 2
Don t make major changes while at competition, a lot of the times that messes the robot up. We also figured out how to get heading hold and dead reckoning to work. We learned how to do basic vision processing. The vision processing we were doing was trying to find the starting gate and the buoys we would have to bump into C. What skills can we transfer from RoboSub to RoboBoat We can transfer the vision processing skills that we got from Robosub because RoboBoat is above the water. Being above the water makes it easier for you to see with a camera. We can transfer over our dead reckoning skills that we got from not being able to use GPS. II. ROBOBOAT STRATEGY We are a team of two people so we won t be able to work on all of the tasks. Because of this we decided to prioritize the tasks[5]. 1. Navigation test (Speed gate, mandatory) 2. Obstacle course, 3. Interoperability challenge 4. Automated docking 5. Locate sonar pinger What excited us the most about the RoboBoat competition was the Interoperability challenge. So we began working on it in January, by February we had had many autonomous take-offs, missions, and landings. We were working on improving the autonomous landing accuracy using vision. A short video of this can be found at [https://youtu.be/bx1v_lactrc] As seen in the video, our vision processing often failed to detect the target and our navigation failed to land the quadcopter exactly on the target. At this point, RoboBoat hang-outs indicated that the flying portion of the Interoperability challenge may not occur. So we have ceased working on this portion of the project. We will be trying to keep costs low by using as much of the submarine from last year as we can. We will also waterproof everything just in case something bad happens. We will be re-using two of the three thrusters from last year. A. RoboBoat tasks To complete the navigation test we have to go through a set of gates. Which requires navigating to a waypoint and using sonar to make sure we don t run into the gates. To complete the obstacle course we need to enter through 1 of 3 gates, the number will be given to us at run time. We then go through the obstacle course, then we exit through 1 of 3 gates which are also told to us at the beginning of the run time. To complete the interoperability challenge the quadrotor has to take off from the boat, fly over to a designated area with a target on the ground, take a picture of the target on the ground, send the picture to the judges, identify and tell the judges what letter is seen on the target, and then finally come back and land on the boat. To complete the automated docking challenge we need to use vision processing to find what dock has the correct color and shape, then dock at that dock, next you back out of the dock and find the next target and dock at it To complete the sonar pinger challenge we will have to listen for the acoustic of the pinger and find the approximate position of the pinger. Then we will have to do a full circle around the pinger B. Implementation options For navigation we considered two major options: APM Drones solutions[6] and custom code in the Vectornav-300[7] 1) APM Drones The APM drones has several implementations such as: APM Pilot, APM Rover[8], APM Copter, but it doesn t have an APM Boat. It should be possible to make the APM Rover steer as a tank that would be the same as the differential steering of our boat. Advantage Navigation and waypoint code is already written and debugged APM Rover code appears to support obstacle avoidance Telemetry links displayed in a mission planner GUI 3
Figure 2 shows the bottom of the boat, sonars, motors, and the electronics boxes. Disadvantage Not as accurate as the Vectornav Interfacing with the RoboBoat judges is not directly supported and will require custom coding 2) Vectornav-300 compass and GPS The Vectornav-300 compass and GPS was donated to us by Vectornav Corporation. Advantage Disadvantage Very accurate positioning We have to write custom code for navigation and motor control. It is very good at giving headings and keeping magnetic anomalies out of our readings. It uses differential GPS to determine heading We didn t use it because the inertial navigation system requires significant motion to lock, and our slow boat didn t seem to be able to move quickly enough for it to lock on. III. S.S. MINNOW HARDWARE A side view of the boat with the quadrotor mounted on top is shown is figure 1 below. Figure 2 The viewed from the bottom The hardware consists of the following components: Compass and GPS - ublox M8N[9] GPS and Compass A 32-bit inertial navigation system supported by Ardurover called a Pixhawk[10] Figure 1 viewed from the side 4
Motor controllers underneath Motor controllers Quicrun 1060 60-amp speed controllers[11] Emergency stop system E-stop The emergency stop shuts the motors down using a relay board. The motor controller is inside of the electronics box. Hidden under the connector blocks. 5
Obstacle avoidance sonar - 2 low cost ultrasonic sensors[13] and an Arduino Uno to convert the signals to analog format to be read by the Pixhawk Relay board Sonar sensors Remote controller Flysky 6-channel 2.4 GHz transmitter and receiver for manual control[12] Telemetry and control link 915 MHz Telemetry link Camera Ethernet camera TP-SC 2020N[14] Flysky receiver Camera 6
Wireless router TP link 841[15] Battery 12V lead acid, 6 Amp hours Router Battery Ground station laptop HP Pavilion G6 Laptop The ground station will be used for monitoring and taking over control of the boat if needed. Mission planner software (shown below) displays the attitude, speed, and positions of the vehicle. It also displays the vehicles progress towards waypoints, battery status, and current GPS position. It can also arm/disarm the robot as well as change flight modes between automatic and manual. Thruster 800 GPH Bilge pump motors[16] Thrusters 7
Interoperability platform Quanum Nova quadcopter, Logitech webcam, and UDOO processor board experience with it and there is already vision processing software written for this board from last year s RoboSub competition. The disadvantage of this board is because its processing capabilities are limited. Our other option is a laptop computer doing the processing. The advantage of this is that there is more processing power and storage. A disadvantage is that it is heavy, expensive, and requires new software. Testing between these two options is ongoing with both seeming promising. Sonar detectors We have some audio microphones and amplifiers, we may be able to detect the pinger location based on volume only. Main on-board computer This computer will control the communications with the judges and control which task the boat is executing. At this point in time is unclear what computer platform we will use. One option is the UDOO computer board, which is a combination of a Linux computer and an Arduino. It s easier to work with because we already have IV. VISION PROCESSING The vision processing in RoboBoat is very similar to RoboSub. We use Python to call on OpenCV to analyze the images on our Linux computer. RoboBoat requires detecting objects in a cluttered environment. Since we don t have a lot of images from RoboBoat we will use pictures from Robosub to show the software detecting the gates. Notice that the gates are not obviously visible in the original image, we hope that this will help indicate that we can detect the gates and targets at RoboBoat. Image Comment Original Image 8
Red Plane Only Equalize Histogram Adaptive Threshold Remove Small Objects Identify Potential Gates Highlight Center of Gate 9
A key application of the vision will be the Interoperability challenge. We will be using vision to detect the target on the ground. We will use blob detection to locate the elements and positions of the blobs to find what character is being displayed on the target. V. TECHNICAL CONCERNS Because of our experience in RoboSub we are pretty sure we can do the obstacle course, the navigation challenge, and the vision processing. In RoboSub you can t communicate with the vehicle while it is underwater. Our concern is that the communications with the judges is something we don t have experience with and it is something that we have limited ability to test. VI. CONCLUSION We have experience with RoboSub. Because of our experience with RoboSub we feel confident that we will be able to go through navigation gate, obstacle course, and the vision processing portion of the static interoperability challenge. We think we had a good chance at the full interoperability challenge if they had let us fly, but landing on the boat was not fully solved. We have promising results for the docking challenge, but no way to test it. We have some ideas for the sonar challenge, but no progress yet. We are only a team of two and we run on a tight budget so we probably will not be able to complete all of the competition challenges in our first year at this competition. We are looking forward to the challenges of RoboBoat 2015. VII. REFERENCES [1] http://www.seaperch.org/seaperch_challenge [2]http://www.auvsifoundation.org/foundation/competition s/robosub/ [3]http://www.auvsifoundation.org/competitions/robosub/p ast-robosub-competitions/2013-robosub [4]https://higherlogicdownload.s3.amazonaws.com/AUVSI /fb9a8da0-2ac8-42d1-a11ed58c1e158347/uploadedfiles/robosub%20final%20scores% 20Ranking.pdf [5]https://s3.amazonaws.com/com.felixpageau.RoboBoat/R oboboat_2015_final_rules_20150527.pdf [6] http://diydrones.com/ [7] http://www.vectornav.com/products/vn-300 [8] http://rover.ardupilot.com/ [10]https://store.3drobotics.com/products/3dr-pixhawk [11]http://www.hobbywing.com/product_show.asp?id=300 [9]http://www.u-blox.com/de/gps-modules/pvtmodules/neo-m8-series-concurrent-gnss-modules.html [12]http://www.amazon.com/Digital-Proportional-Channel- Transmitter-Receiver/dp/9269807304 [13]https://docs.google.com/document/d/1YyZnNhMYy7rwhAgyL_pfa39RsB-x2qR4vP8saG73rE/edit [14]http://www.tplink.com/lk/products/details/?model=TL-SC2020N [15]http://www.tplink.com/lk/products/details/?model=TL-WR841ND [16]http://www.iboats.com/800-Gph-Bilge-Pump-24v- Round/dm/cart_id.395235083--session_id.432756595-- view_id.1168992 10