March Progress Update

Hardware Team:

This month, the hardware team tackled some structural changes to the overall frame design of the drone as well as implemented a new design for our tank, which will hold the pesticide while spraying, and spray boom.

Additionally, to house the brains of the drone, we needed some sort of cover. Normal dome covers are available but are not up to par when it comes to the structural integrity that we want in case of a situation where sensitive parts of the drone are at risk.

To alleviate the issue, we custom manufactured a dome that will act as a permanent housing with a removable top that will lock and lift when needed. The dome itself is being 3D printed by the Middlesex County College.

For spray boom design, we established that the boom needs to be flexible and have wheels to allow for proper landing. We have a few different designs which we have ordered and received. Now, we are in the process of testing the different designs to see which will best suit our needs.

two wheels in between 2 nozzles design

two wheels in between 2 nozzles design

1 wheel in between 2 nozzle design

1 wheel in between 2 nozzle design

Navigation Control Team + Hardware Team + User Interface Team:

At this point, many of our sub-teams have begun integration and are now working with each other. With the drone built, the next major milestone that the Navigation Team, along with the Hardware Team and User Interface Team, has been working on is to have the drone fly stably under control of a joystick. Through various tests and examining the data recorded by Mission Planner, we found out that one of the reasons why our drone did not always respond how we wanted it to was due to the calibration. We re-calibrated the drone after we replaced parts such as arms and motors. As a result, our drone responded more accurately.

One thing that we learned through our flying tests with the drone was that it is not easy to control and navigate. Even with a joystick, the driver must learn to transition smoothly and not move the shaft with sudden force. This required practice, and although we damaged a few parts in the process, we are now able to successfully control the drone with the joystick!

After this milestone, the next is to master autonomous flight using the Mission Planner control software. The UI team, in preparation, have been adding finishing touches on their
interface.  The team added a way of outlining a field on the map.

UI

Whenever a field is created, the user has to outline the field. It is then saved and can be displayed whenever it is selected from the field list.

Image Processing Team:

While flight testing is ongoing, the image processing team is working on refining their algorithms. We have two different algorithms, one is a contingency plan for the other. The more complicated algorithm involves texture detection and a complicated form of mapping. The contingency plan involves color detection and a color-based mapping algorithm. The latter is concerned with a very specific farm field whereas the former is advanced enough to apply to nearly all crop geometries.

Image processing, as well as the pesticide spraying mechanism and tank, will be integrated into the main system soon after we master autonomous flight using Mission Planner. Meanwhile, we are preparing by working on integrating the UI team with the IP team.

Posted in Uncategorized | Leave a comment

Guidance Day

This past Wednesday was Guidance Day at our school. In celebration, our InvenTeam presented our drone during a meeting attended by regional guidance counselors. Their support and enthusiasm for our project was wonderful to see and a fun experience. Check out the pictures below!

This slideshow requires JavaScript.

Posted in Uncategorized | Leave a comment

Flight Testing and Failure Analysis

On March 16, we took to the skies for our first unrestricted flight test! In previous tests, we kept the drone tethered to the ground but now, the training wheels were off.

Of course, a day of testing wouldn’t be complete without a crash! And that is exactly what happened. In the fifth round of testing, the drone experienced a failure. It fell on its side during a flight test using the joystick.

Fortunately, Mission Planner has flight logs which we were able to look back to. These logs document the drone’s specifics and the joystick control’s levels during each flight. Failure analysis indicates that the drone lost contact with the joystick when the throttle was changed from up to down. This can be attributed to the drone’s calibration being slightly off, resulting in a slight tilt which was overcompensated for by the throttle.

Roll and pitch throttle input from joystick

The images on the right show some of the data retrieved from the flight log. The top image shows the pitch and roll of the drone at the time of the crash. You can see that the pitch (red) and the roll (green) dramatically changed just before the crash. The bottom picture shows the throttle input received from the joystick. The throttle rises significantly but instead of gradually declining, the throttle dramatically plummets, which leads us to believe this is where the drone lost connection to Mission Planner.

The result? A fractured arm. Although no propellers were damaged, one of the arms sustained injury enough to render it unusable.

IMG_2956

Proper propeller + arm arrangement

IMG_2955

Damaged arm after crash; arm no longer allows for free propeller movemen

 

 

 

 

 

 

 

The arm has been replaced with a spare and we will be up and flying again this Friday. This time, we have experience and knowledge to prevent such a failure.

Posted in Uncategorized | Leave a comment

Obstacle Detection Using Sonar

The last thing we want our drone to do is crash into something in its path. How do we prevent this? Obstacle detection. However, image processing for obstacle detection may not be entirely effective. It is incredibly difficult to account for all potential obstacles with a high level of accuracy due to all the changes that occur in lighting and color throughout the day. If obstacle detection in the code fails us, we needed a fallback plan to prevent potential damage to the drone.

sonarThis is where sonar comes in. We found a sonar sensor that is compatible with the ArduPilot. Mounted on the spray boom, this will increase our ability to detect obstacles as we will now have two methods in case one fails us.

 

Posted in Uncategorized | Leave a comment

Mid-Grant Technical Review Overview

After countless hours of preparation, our hard work finally paid off! The Mid-Grant Technical Review was a success and we learned a lot as a team. Various local farmers attended and provided us with helpful advice for us to keep in mind while we continue building our drone.

Team after the MGTR Presentation

Team after the MGTR Presentation

IMG_0930
For the presentation, the Navigation Control team, in conjunction with the Hardware team demonstrated the drone in quadrotor formation, with 4 motors powered by 4 batteries, controlled by the transmitter. The team demonstrated the different motor commands, such as yaw, pitch, roll, and throttle, through the controller.

IMG_0890

The Image Processing team presented their green detection code with a live feed which was displayed on the smart board. The team talked about the organization of the code and their camera system.

The User Interface team presented the functions of their version of the Mission Planner program.

And lets not forget the awesome demo of the nozzle tester provided by TeeJet Technologies.

IMG_0912For more pictures of the event, be sure to check out our gallery as well as our video below!

Once again, thank you to everyone who came out to support us!

Posted in Uncategorized | Leave a comment

February Progress Update

Navigation Control Team:

This month, the Navigation Control Team had their first flight test. The stability/flight test was done using the joystick to lift the quadrotor drone a foot in the air. The drone tethered to the ground for safety purposes.

The Navigation team is now working closely with the hardware team to be able to control the drone’s full range of movement as an octorotor instead of a quadrotor.

Hardware Team:

At our Mid-Grant Technical Review, the team learned that most pesticide formulas require an agitation supplement to prevent the pesticide mixture from settling in the tank during idle periods. A suggestion made by Mr. Jany, a Christmas Tree farmer, was to implement an agitation method via the demand pump for several reasons.

His idea was to have a secondary output emit from the pump, allowing it to constantly feed back into the tank when not spraying to allow the mixture to be agitated thoroughly. This “kills two birds with one stone” as it limits the time the demand pump will need to start up and shut off.

Overall, this addition will allow us to keep the consistency of the pesticide to an ideal level and limit vibrational movements from the pump.

___________________________

In addition to the the tank redesign, the frame has been fully assembled to its octocopter form. After some initial flight testing with the landing gears of the drone tied down, it is time to retest with all eight motors and arms.

Octocopter in Progress

Octocopter in Progress

User Interface Team:

A major change in functionality that we have started is looking into communication between the Image Processing and the drone itself. We made it possible for Mission Planner to now read from a file every few seconds to check for updated directions given by the image processing. Mission Planner then sends the direction instructions to the drone. For example, if the instruction in the file is “right”, a command to turn the drone right will be sent. The process isn’t entirely functional yet, but it is nearly there.

shows the command "left" being read from the output file created by the image processing

shows the command “left” being read from the output file created by the image processing

An alternative to reading from a file that we have started looking into is communicating directly between the Mission Planner program and the Image Processing program. Direct communication between the two programs would get rid of any latency incurred from having to read from a file, which is very important. The difficulty lies in the fact that Mission Planner is written in C# and the Image Processing is done using the Open CV library which is written in C++. We have started researching about how to make this communication happen, and we will try to see if it will work.

 Image Processing Team:

This month, the Image Processing team began looking further into obstacle detection. We were able to detect rows of crops using green color detection, however, this limited our capabilities in detecting obstacles. To overcome this, we decided to detect the color brown instead. This way, the ground will be detected and anything out of the ordinary shades of brown will be flagged as obstacles.

Noise in brown color detection

Noise in brown color detection

Unfortunately, the ground also has a substantial amount of gray and various other colors in it, and the obstacles we are looking for (crop alignment sticks) are also brown. Thus, we had to create a different obstacle detection algorithm which we are now in the process of coding. Now, we are attempting to use a combination of green color detection with an overlay of shape detection to be able to identify these rectangular sticks as obstacles while still detecting plants in the path.

Posted in Uncategorized | Leave a comment

Preparing for Mid-Grant Technical Review

Thursday February 27, the MCA InvenTeam will be hosting our Mid-Grant Technical Review. The MGTR is an opportunity for our InvenTeam to present our current progress and future plans to an audience and receive feedback. After hard work and preparations, our team could not be more ready and excited for this event.

This slideshow requires JavaScript.

For more information on the event, please see our electronic invitation or contact us.

Posted in Uncategorized | Leave a comment