Please visit the respective links to see the progress of each subsystem.<\/p>\n
\n
<\/a><\/a>Integrated System<\/strong> Finished integration of the system. Here is the link of the final demo video demonstrating the functionality of the\u00a0integrated\u00a0system – link<\/a><\/p>\n The overall system architecture and the UAV with all the components attached are shown below<\/p>\n April 13, 2016<\/strong><\/p>\n Changes in obstacle avoidance system<\/strong><\/p>\n We discovered our obstacle avoidance sensor \u00ad Hokuyo URG-04lx generates noisy data in outdoor environment. Hence we implemented the following solutions:<\/p>\n The above figure shows the rviz \u00a0visualization of the UAV avoiding obstacles after replacing the sensor and applying the solutions mentioned above<\/p>\n March 31, 2016<\/strong><\/p>\n Package delivery without obstacles<\/strong><\/p>\n The UAV is able to deliver packages using the Electro-Permanent Magnet. The package upto 100g can be delivered to within 2m distance from the marker.<\/p>\n https:\/\/youtu.be\/qiPCzdjwa_0<\/p>\n In the video above, the UAV takes off from starting point, ascends to 15m height, and goes to the pre-defined house position. Then, the UAV searches for the marker using a lawnmower search pattern. Once the marker is found, the UAV descends onto the marker, lands, and drops the package.<\/p>\n Lawn Mover search implementation with obstacle avoidance<\/b><\/p>\n We have achieved the functionality of the UAV traversing over\u00a0a lawn mover search path to find the marker while avoiding obstacles.\u00a0<\/span>We flew the UAV at around 15m height the UAV followed the lawn mover path well.\u00a0As we don’t have obstacles that high we could not verify the obstacle avoidance subsystem at that height. Next we will test obstacle avoidance using the test obstacles that we build for our project.<\/p>\n Physical Obstacle Design and Fabrication<\/strong><\/p>\n The test requirements for the drone required that our project required the\u00a0obstacles to be 1.5 x .5 m and 2 x 2 m. Upon further discussion as a team, we determined\u00a0that two obstacles of 1.5 x .5 m would be more financially responsible in addition to\u00a0raising the standards of the project as smaller objects are harder to detect. Original Obstacle Design<\/p>\n However, upon inspection the \u00be\u201d PVC pipe was too thin to support the structure\u00a0and would possibly buckle in the wind. 1-\u00bd\u201d PVC would suffice but would also double the\u00a0price.\u00a0The team settled on a second design that was significantly less costly. The\u00a0second design which is shown below.<\/p>\n Updated Obstacle Design<\/p>\n The new design is significantly cheaper and easier to transport. It uses two 2\u201d PVC\u00a0for a support shaft. Tarp mounted to a 2×4 wood frame provides the body of the \u00a0obstacle. The body will be hoisted into place by rope like a flagpole.<\/p>\n Tarp over wooden frame<\/p><\/div>\n Close-up of Frame with Tarp Attached<\/p><\/div>\n Obstacle Base<\/p><\/div>\n Obstacle partially erected to 10ft due to space limitations in lab<\/p><\/div>\n Fabrication of the final design took two days. We identified several improvements during the build that were incorporated into the final design. The hardest part of the build was the wooden frames for the tarp. Braces extend from all four sides to buttress the pole and prevent movement.<\/p>\n Updated Software Architecture<\/strong><\/p>\n Manual RC control of Electro Permanent Magnet<\/strong><\/p>\n https:\/\/youtu.be\/KhNbaFDekIE<\/p>\n <\/p>\n March 17, 2016<\/strong><\/p>\n Finishing Obstacle Avoidance code integration with Pixhawk- Running Obstacle Avoidance code off Odroid<\/strong><\/p>\n https:\/\/www.youtube.com\/watch?v=r1X6UWgTfsM<\/p>\n Above shown\u00a0is a video of the UAV performing\u00a0autonomous navigation to the goal position while avoiding obstacles. \u00a0<\/span>The UAV can detect an obstacle from a distance of 3m and it maintains a distance of 0.5m from the obstacle at all <\/span>time. <\/span><\/p>\n Autonomous Takeoff and Landing<\/strong><\/p>\n https:\/\/www.youtube.com\/watch?v=Jz0qRndWkbM<\/p>\n In the above video, autonomous takeoff and landing functionality is demonstrated.<\/p>\n <\/p>\n February 25, 2016<\/strong><\/p>\n Primary integration of obstacle avoidance code with Pixhawk flight control code<\/strong><\/p>\n We have achieved communication between Navigation Stack and\u00a0the Pixhawk. After testing the code in Software In Loop Simulation of Pixhawk (as mentioned in Obstacle Avoidance Subsystem) we tested the code functioning on X8+.<\/p>\n Integrating Navigation Stack obstacle avoidance code with Pixhawk code<\/p><\/div>\n The output of odometry and navigation stack changed as\u00a0we\u00a0moved the UAV around manually. This verified that the Navigation Stack and the Pixhawk are communicating as needed. \u00a0<\/span><\/p>\n <\/p>\n February 10, 2016<\/strong><\/p>\n Autonomous Navigation from A to B using Guided mode<\/p>\n https:\/\/www.youtube.com\/watch?v=psoAnLxfEpo<\/p>\n The UAV takes off using manual RC control, then,\u00a0the mode is switched for autonomous navigation. The UAV stays at the original point\u00a0for 2 seconds and then moves ahead (UAV X axis) by 4m, and increases its altitude by\u00a01m. It then holds its position until mode is switched back and the UAV is manually\u00a0landed.<\/p>\n Autonomous Hover over Marker<\/strong><\/p>\n <\/p>\n https:\/\/www.youtube.com\/watch?v=x94lj6qUGSQ<\/p>\n As can be seen in the video, after the UAV is put into\u00a0autonomous mode (audible beep), the UAV positions itself over the marker and\u00a0maintains position at 18m height until its taken back to manual mode and landed.<\/p>\n The nested AprilTag marker is detected from the UAV and its target position is updated continuously to maintain position over the marker<\/p>\n ROS architecture for autonomous hover over marker of the UAV. An apriltag detection node publishes the position of marker with respect to UAV. A setpoint publisher uses that and position data of the UAV to publish setpoints in global frame. (Takeoff point is origin)<\/p><\/div>\n <\/a> April 20, 2016<\/strong><\/p>\n After February end we started integrating obstacle avoidance subsystem with the complete system. The details about the integration are mentioned in the integration section. The flow chart shown below describes the final logic that was used for obstacle avoidance<\/p>\n February 25, 2016<\/strong><\/p>\n Testing Navigation Stack functionality with Software In The Loop Simulation of the Pixhawk. We could pass odometry data from UAV to navigation stack. Receiving this data the navigation stack passed velocity commands to UAV to move towards goal.<\/p>\n In the above figure, the tab on the upper left is the console which shows the data about the status of \u00a0the UAV (its altitude etc.). The bottom left figure is the SITL map showing the virtual UAV in a virtual map. On the right is the rviz visualization of the UAV. The red arrow are the location of UAV. As seen in the figure the continuous vertical red line shows the UAVs takeoff. The green line is the path generated by navigation stack to reach the goal position (shown by a single sed line on the surface). <\/span><\/p>\n <\/p>\n February 10, 2016<\/strong><\/p>\n In simulation got the robot to\u00a0follow the path planned by the planner.<\/p>\n Here is a video\u00a0showing lawn-mover search implemented in simulation.<\/p>\n https:\/\/www.youtube.com\/watch?v=RDcBtPeICZ8<\/p>\n January 28 2016<\/strong><\/p>\n We redid the trade study for Obstacle Avoidance Subsystem sensors as we shift to X8+ UAV platform. Now we are going to use LIDAR for obstacle avoidance sub-system. We are using Navigation Stack of Robotic Operating System for implementing obstacle avoidance.<\/p>\n The navigation stack contains the following parts: <\/p>\n Video below shows basic test of LIDAR interfaced with\u00a0Navigation Stack of Robotic Operating System<\/strong><\/p>\n https:\/\/youtu.be\/iO8Klr5yQxE<\/p>\n <\/p>\n We are using LIDAR instead of using 14 ultrasonic sensors because of the major reasons:<\/p>\n <\/p>\n December 18 2015<\/strong><\/p>\n The video below shows the functioning\u00a0of 6 ultrasonic sensors mounted on the nose of the FireFly6 UAV<\/p>\n https:\/\/www.youtube.com\/watch?v=fZqf8UesqAo<\/p>\n December 7, 2015 – Fall Validation Experiment<\/strong><\/p>\n Figure below shows the 6 ultrasonic sensors mounted on the UAV.<\/p>\n <\/p>\n Close Up of 6 Ultrasonic Sensors Mounted at the Nose of UAV<\/p><\/div>\n Image Showing 6 Ultrasonic Sensors Mounted at the Nose of the UAV<\/p><\/div>\n Flowchart of our obstacle avoidance system<\/p><\/div>\n <\/p>\n Master-Slave Sensor Boards<\/strong><\/p>\n As mentioned in the flowchart above\u00a0we are using I2C communication for connecting 14 ultrasonic sensors to the flight controller. Hence we designed master board to handle combining of I2C lines and 5V regulator. Slave boards were designed to take sensor inputs and reduce analog line noise by converting it straight to digital.<\/p>\n Flowchart of connection of sensors with the master-slave board<\/p><\/div>\n Fall validation master-slave test setup<\/p><\/div>\n November 07, 2015<\/strong><\/p>\n CAD model of ultrasonic sensor layout on UAV<\/p>\n UAV with 14 ultrasonic sensor around it<\/p><\/div>\n <\/p>\n Top view of sensor layout<\/p><\/div>\n Front view of sensor layout<\/p><\/div>\n The ultrasonic sensors are to be pinged serially to avoid interference among sensors. Each sensor takes around 50ms. So pining 14 sensors one after the other will take 50 X 14 ms = 700ms which is too much. SO we further divided sensors into 3 groups as shown below.<\/p>\n Subdividing the sensors into 3 subsystems to reduce uodate rate of iverall obstacle detection system<\/p><\/div>\n Master-Slave Sensor Boards<\/strong><\/p>\n We are designing master-slave I2C board to connect all the sensors to the flight controller<\/p>\n Two Printed Slave Boards<\/p><\/div>\n <\/p>\n Schematic of 4 slave boards<\/p><\/div>\n Schematic of the Master Board<\/p><\/div>\n October 30, 2015<\/strong><\/p>\n Following are the tasks accomplished until now:<\/em><\/p>\n
\nApril 20, 2016<\/strong><\/p>\n![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Screenshot-from-2016-05-06-193944.png\")
<\/a><\/p>\n![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Screenshot-from-2016-05-06-194039.png\")
<\/a><\/p>\n\n
![\"New](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/lidar.png\")
<\/a><\/p>\n![\"Obstacle](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/obs_avoid.png\")
<\/a><\/p>\n
\nThe original design for the obstacles required that these profiles be elevated two\u00a0stories in the air (approximately 6 meters). I planned to implement these using a\u00a0scaffolding of \u00be\u201d PVC pipe. Tarp would cover the scaffolding to provide a surface for the\u00a0Hokuyo Lidar to detect.<\/p>\n![\"Snapchat-5551699698106715790[1]\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Snapchat-55516996981067157901-169x300.jpg\")
<\/a><\/p>\n![\"Sketch](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Snapchat-22495738970451195351-e1459559865399-300x234.jpg\")
<\/a>![\"Obstacle](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/ObstacleMockup-e1459560148270-154x300.png\")
<\/a><\/p>\n![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/SAM_1897-300x200.jpg\")
<\/a>![\"SAMSUNG](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/SAM_1899-300x200.jpg\")
<\/a>![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/SAM_1900-300x200.jpg\")
<\/a>![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/SAM_1896-644x1024.jpg\")
<\/a>![\"Control](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Control-Architecture-v2.png\")
<\/a><\/p>\n![\"Integrating](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/pic2-1024x632.png\")
<\/a>![\"Node](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Node-architecture-auto-flight.jpg\")
<\/a>
\nThe Obstacle Avoidance Subsystem<\/strong><\/p>\n![\"\"](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Screenshot-from-2016-05-06-185936.png\")
<\/a><\/p>\n![\"SITL](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/pic1-1024x656.png\")
<\/a><\/p>\n
\nGlobal costmap
\nGlobal path planner \u2013 It uses the global costmap to compute paths ignoring the kinematic\u00a0and dynamic vehicle constraints. It uses Dijkstra\u2019s algorithm to do this.
\nLocal costmap
\nLocal path planner \u2013 It accounts for the kinematic and dynamic vehicle constraints and\u00a0generates feasible local trajectories in real time while avoiding obstacles
\nMove_base \u2013 It implements the state machine<\/p>\n\n
![\"Close](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/close_up_view_sensor_mount.jpg\")
<\/a>![\"Image](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Sensors_mounted_on_UAV.jpg\")
<\/a>![\"Flowchart](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Flowchart_obstacle_avoidance.png\")
<\/a>![\"Flowchart](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/pcb_flowchart.png\")
<\/a>![\"Fall](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/fall-validation-sensor-board.png\")
<\/a>![\"UAV](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/09\/Graphical-representation-of-uav.jpg\")
<\/a>![\"Top](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/top_view_sensor_cad-1024x587.jpg\")
<\/a>![\"Front](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/front_view_sensor_cad-1024x448.jpg\")
<\/a>![\"Subdividing](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/sensor_subsystems.png\")
<\/a>![\"Two](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Two-Printed-Slave-Boards.jpg\")
<\/a>![\"Schematic](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/SlaveBoards.png\")
<\/a>![\"Schematic](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Master-Board.png\")
<\/a>\n
![\"Sequential](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/multiple_sensor_pinging.png\")
<\/a>![\"Sensor](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/IRUSsensor.png\")
<\/a>![\"Obstacle](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/10\/Obstacle-Avoidance-Subsystems-v1.jpg\")
<\/a>![\"Flowchart](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/flowchart.png\")
<\/a><\/p>\n![\"Error](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/error_graph.png\")
<\/a>![\"Nested](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/tag_comb_rot_54cm-300x300.png\")
<\/a>![\"Nested](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Screenshot24-1024x576.png\")
<\/a>![\"Nested](\"http:\/\/mrsdprojects.ri.cmu.edu\/2015teama\/wp-content\/uploads\/sites\/2\/2015\/12\/Screenshot26-1024x576.png\")
<\/a>