Engineering a Modular Solution to the Future of Transportation: Drones
Advisor: Dr Sanhita Dixit
Internship Location: Service Now
Onsite Mentor: Abhishek Kumar
Package delivery systems encounter the "last-mile" delivery challenge, which entails efficiently transporting goods from a distribution center to the final destination, typically a residential area, while overcoming logistical complexities and minimizing costs. Current solutions rely on fossil-fuel-powered vans that contribute to pollution and worsen traffic. Enter drones, a sustainable and efficient mode of transportation with the advantages of being battery-powered and using straight-line routes to destinations instead of roads like conventional delivery vans. Most packages are small and lightweight, well within the scope of medium-sized drone payload capability. My project aims to address the current limitations of delivery drones, characterized by excessive noise, inefficiency, and inadequate landing capabilities. My methodology involves rigorous lab work to determine optimal propeller setups, experimentation with adjustable pitch propellers, and fabricating lightweight drone components to support the payload. With the end product, I aim to showcase a fully functional drone with variable pitch, obstacle avoidance, payload handling, winged flight control, and battery swap technology, demonstrating a holistic approach to drone innovation.
Project Blog
Week 0: It's a Bird... It's a Plane... It's a Delivery Drone!
Whoosh, bzzzzzzz, ding!
Welcome to my delivery drone development blog, where I'll take you on an adventure to discover the possibilities of tackling the last-mile delivery challenge for deliverable goods using drones.
"What is the last-mile delivery challenge?" you may ask. Well, it's the final stretch of the delivery process, where goods are transported from a distribution center or hub to their ultimate destination, often a residence or business. This phase is typically the most complex and expensive part of the delivery process, as it involves navigating dense urban environments, dealing with traffic congestion, and ensuring timely and accurate delivery to customers' doorsteps.
Many have proposed drone delivery to replace the diesel trucks that roam our streets. And the idea does have many merits! Drones' linear paths significantly improve route efficiency and reduce the overall carbon footprint of delivery operations. Drones also provide greater flexibility and scalability as they can reach locations that may be inaccessible or difficult for traditional delivery vehicles to access, such as remote rural areas or dense urban neighborhoods with narrow streets.
However, current drone delivery solutions are not ready to tackle the last mile. They are too loud, inefficient, and unintelligent. Some challenges to innovate include precise drop-off locators, quiet propeller design, battery swap capability, and adequate payload capacity. In my senior project, I aim to develop a system that does just that.
So, why me?
I've had a lifelong passion for transportation, from trains to planes. My curiosity naturally shifted from RC cars to UAVs as I grew older. With years of advanced physics, mathematics, and design/engineering expertise, I am ready to innovate and push this field forward.
To build my autonomous delivery drone platform, I'm working with Mr. Abishek Kumar, who holds multiple patents in drone technology, and Dr. Sanhita Dixit, my physics teacher since 7th grade and internal advisor.
If all goes well, expect a fully functional, autonomous drone that solves most of the problems I mentioned above by the end of the school year! Cheers to a wonderful third trimester filled with creative thinking, design challenges, and burnt PCBs.
Aidan Peter, out.
Week 1: Variable-Pitch Propulsion, Like Thrust Vectoring for Drones
Hey, it’s you again! Welcome back to my delivery drone development blog.
When I haven’t been building, I’ve been brainstorming names for this project. Deliverone, FlyDrop, SkyAge. Ehh, not great. What about Hüver? I like that.
I’m sure you’re wondering what else I’ve been doing besides coming up with names for the project. Well, I’ve been developing a test stand for my variable-pitch system for drone propulsion.
What is variable-pitch? It’s a system that is employed on many modern helicopters that allows the pilot to adjust the pitch angle of the propeller. When pitch angle is increased, each blade sweeps up more air and produces more lift. Think about it like a bicycle gear system. On a high gear, you have to pedal harder for each rotation but you travel further. For systems like this, there’s often an optimal cadence where certain gears are most efficient.
Applying this technology onto drones can revolutionize stability and maneuverability by speeding up reaction times for motor speed adjustments. On typical drones, the systems rely on the moment of inertia of the motor to slow down the motor. With a variable pitch system, we use an electronic servo to adjust the pitch of the propeller which results in fine adjustment capability that drastically outpaces the traditional acceleration and deceleration of motor rotation speed.
Watch this video from MIT for a better visual on variable-pitch effectiveness: https://youtu.be/Vy5Ky50eGJs?si=s_loo6c8umhMFFzm
Now, building the variable-pitch apparatus has been no easy feat.
It requires an out-runner motor with a hollow shaft so that the long variable-pitch level can fit through the motor to be adjusted from under it. Here's a picture of how the motor looks like after I replaced the solid shaft with a hollow one:
The pitch is adjusted through a linear actuation which pushes the thin shaft up and down to rotate the propellers with ball bearings. For demonstration purposes, I will be using a servo attached to a rack and pinion to push and pull for pitch adjustment. Here's what that looks like:
Next, I put it all together. Using a long, 3D printed stand and a zip tie to attach the rack to the push-pull lever on the variable-pitch apparatus, it all successfully came together to create a servo driven variable-pitch propulsion prototype! (That's quite the tongue-twister isn't it?)
So... what comes next?
I'm having some trouble attaching the propeller to the motor, but that issue should be resolved shortly. Once that's figured out, I'll power everything on. The motor runs on an 11.1 V LiPo battery and the servo will connect to an Arduino board so that I can calibrate the pitch actuation. Once this works in tandem, I want to measure the change in thrust with various pitch angles while keeping the throttle constant by using a load cell. I'll even measure the RPM using a laser tachometer to see how much the props slow down when sweeping more air (conservation of energy!). I'll graph all of these quantities to give you a nice visual.
That's all I have for you this week! Check back in next week for updates on Hüver. I'll be starting work on a sustainable landing pad design so that the drone can easily locate the drop-off target.
Week 2: Graphs, Cameras, and Design!
Has it already been another week? Time flies! Now, if only my drone could... I'm working on it.
Anyways, welcome, reader, to your weekly installment of delivery drone development! Last week, I code-named the project "Hüver" and began construction of a variable-pitch test stand. Initially, I prototyped with a rack and pinion mechanism (see previous blog for further detail), but at the tail-end of last week, I decided to switch to pure servo control. This decision stemmed from the rotational nature of the servo, which doesn't deviate the path of the shaft enough to create significant flex or damage.
I know that sounds pretty abstract. Here is an image to help you visualize what I mean:
After engineering this system, I finally powered it on. I connected the servo to a Raspberry Pi and added a joystick to the circuit so that I could control the pitch on the fly (get it?!). How satisfied I was when it all came to life after a few short lines of code in the Arduino IDE!
Oh, if only it were so simple. The pitch is directly controlled by the shaft, which pushes up and down, and in turn, is controlled by the rotational position of the servo. The issue with connecting these systems directly is that in the case of a violent jerk or large movement by the servo, the whole system will bend or break. With a small bug in my code, I did just that, and the shaft along with the hook that attaches to the servo arm split in half. Thank goodness for spare parts.
I assembled the mechanism again, and this time gave special attention to my code, strictly limiting the range of motion of the servo using software. I've always been more of a hardware guy, but it's moments like this that help me appreciate the synergy of software and hardware to create a functional product.
Next, I 3D printed a mount for the variable-pitch apparatus so that it can sit atop an HX711 load cell, which is screwed into a wood block. This load cell measures the thrust produced by the motor at various pitch angles and throttle settings. Operating an HX711 load cell involves wiring it to your Arduino and then calibrating it. By default, it produces arbitrary numerical values that correspond to the weight of the object on the cell. You must use objects with known mass to find the appropriate scale so that the device yields measurements in units that we are familiar with, like grams, pounds, or megagrams.
Fun fact: I calibrated the load cell using an iPod Nano, a 100 Yen coin, and a 1/4-inch bolt.
Now, it was time to gather data! Using a digital tachometer, the load cell, and Arduino input, I graphed RPM, thrust, and servo position, which corresponds to a yet-to-be-measured pitch angle. I plan on fine-tuning this data-collection process in the upcoming days, but here are some initial graphs to ogle at: Click this link for my graphs.
There are various improvements to implement. For one, the motor/servo column is not vibrationally isolated at all. The whole system shakes. The servo is not well mounted, so it jostles a bit upon arm movement. Throttle calibration is not optimal as I am using a manual dial. I will switch to Arduino control for this as well, so I can get consistent and repeatable measurements.
Aside from working on variable-pitch, I've been doing a LOT of learning. I enrolled in a course called the Drone Dojo where I am taking courses on drone design, precision landing, and Raspberry Pi integration for autonomous flight. It seems that there is a lot I don't know. I am also learning how to use OpenCV and 3D simulation tools to see how the delivery system will function in the real world. Malcolm X once said that "education is the passport to the future." That appears to be the case here.
I'm currently 3D modeling the chassis for my drone and 3D printing an enclosure for the Raspberry Pi camera that I'll be using for precision landing vision. I'll take many pictures as the drone slowly comes together in the coming weeks.
Finally, I want to give a special shout-out to the zip-ties that hold my every creation together! I truly could not do this without you.
So... see you next week?
Week 3: R&D a.k.a Research and Development
Finally, it's you again! I've been waiting to share more about my drone delivery adventure with you over the last week.
Oh, and what a week it's been!
As you can tell by the blog title, I've been doing a lot of research and development on my drone. I began by sketching away, leveraging years of art classes to render prototype designs for my drone. There are a multitude of factors to consider when building a drone from scratch, like I am. And over this week, I got lost in all of these details.
Here's a list of all the important questions I had to ask when designing the drone (this might not even be all of them!)
How much weight should the drone be able to lift?
What should be the weight of the drone?
What flight controller will be compatible to program with my Raspberry Pi (not edible by the way)?
How much KV should my motor generate?
What must the diameter of that motor's shaft be?
How much current does the motor use at peak operation? What about at idle?
What kind of ESC will be able to handle this amount of current?
What kind of power distribution board can manage that current x 4 (because there are four motors on my drone)?
What kind of batteries will be needed to power this drone for about a half hour?
What kind of propeller should I use to generate the most efficient flight trajectory?
Should it be made of carbon fiber, wood, or plastic?
Will I need to design an adapter so that it can fit on my motor?
What material should I use to build the drone frame?
The questions go on and on and on. A lot of last week has been research into each of these questions, parsing through forums, attending courses, and speaking with personal contacts like my external advisor. I haven't answered all of my questions (and new ones constantly appear), but I've gotten to the point where I can make decisions and proceed. Here's what I've decided.
For the frame, I am using carbon fiber tubes that are joined with 3D printed parts (I have designed). It will use a modular structure to easily expand the storage for electronics and batteries so that in case I something doesn't go as planned, I'll be able to alter the layout. I'm doing this by creating a sandwich-like layout for my drone where there will be electronics, then a layer of 3D printed material, then another set of electronics.
As for the parts, I'm ordering a 320 KV drone motor because of my payload requirements being the average Amazon package (about 5 pounds or so). Using a low KV drone motor for a high payload application is beneficial due to its efficiency, torque, and reduced stress on the motor. This choice will make my drone smoother, with more stable flight characteristics, optimizing the drone for lifting variable-weight payloads while extending flight time and ensuring longevity. I'm still deciding on a propeller, but I'm narrowing in on a carbon fiber set because they are far more rigid and provide much more rigid flight dynamics.
I've ordered all of the compatible electronics as well. I don't want to bore you too much with technical details but I'll be using a peak current of 160 Amps, so I simply bought parts that can withstand that current.
Now, while everything arrives, I'm continuing my education on precision drone delivery. I purchased a course from the Drone Dojo that gives me step by step instructions on how to implement various modules of my project. I'm also using CAD software to design parts that will hold the drone together. I've already begun 3D printing them and running test-fits so that I can redesign flawed parts.
And with that, this week is a wrap! I can't wait for all of my parts to arrive so that I can finally take you all to the skies.
Until next week,
Aidan Peter
Week 4: The Waiting Game
Ah! Hello again. Let's play the waiting game, shall we?
My drone components are still shipping, so I've had to focus on preparing all of the other components so that I'll be ready to fly in no time once the motors, propellers, batteries, and power distribution board arrive at my doorstep.
A large focus of my project is building a drone from the ground up, meaning I need to fabricate a majority of the components myself while keeping cost-savings and weight-savings in mind. Over this week, I've been engineering the hardware that the drone will operate with. I began with a main chassis which has 4 slots for the carbon tubes to fit into. It's structurally reinforced using large blocks with high infill. After 3D printing the version 1 chassis design, I fitted the carbon tubes into it and did some stress-testing. To me, stress-testing simply means putting as much force as I can on each tube while isolating the others to ensure the frame can withstand high forces that may be observed during a crash or rapid acceleration. In the end, the 3D printed chassis broke. Time to go back to the drawing book!
In addition to designing the chassis, I've come up with an innovative system to mount electronics. I call it The Sandwich. It's modular, simple, and space efficient. Basically, it uses a 3D printed plate with spacers which is then mounted to another 3D printed plate. In the space between these plates, there's room for electrical components (like the power distribution board and batteries) with holes for wires to pass through for cable-management.
A huge problem with DIY drones is their instability. Because many parts are custom fabricated, there may not be the structural balance that a tried and tested platform may provide. I must keep this in mind when mounting components to the chassis. Without structural balance, vibrations are to be expected. These vibrations can be critical to the control of the drone as violent movements can confuse the Pixhawk flight controller which uses gyroscopes and accelerometers to stabilize flight. My solution to this is a damped platform for the flight controller. I designed a platform that stands on rubber dampers so that impacts are absorbed before reaching (and confusing) the flight controller.
I learned to do this from a critical mistake on my first DIY drone, which yawed uncontrollably upon takeoff. This week, I also spent some time parting out that drone. I removed the flight controller and electronic speed controllers (ESCs) so that I could use them on my new design.
And with that, Week 4 is a wrap! Thank you for reading, and keep an eye out for next week's blog where I'll make advancements in both the hardware and software to prepare the prototype drone for its first flight.
Week 5: Meet Frisbee!
The waiting game finally ended. My parts arrived. Motors, ESC's, power distribution boards, and batteries dropped from the heavens onto my porch.
Finally, it was time for action.
Over this week, I've realized that building a drone from the ground-up was tough. Very tough. At times, it felt like I was kayaking up a creek without a paddle. But I persisted.
As I mentioned last week, I have to custom fabricate all of my parts. Using my handy dandy 3D printer, I can do this all in-house (or rather in my garage). I finished 3D printing the plates for The Sandwich (refer to last week's blog for context) and they fit together nicely with brass spacers.
Then, I 3D modeled the motor mounts. With the naïveté of a toddler, I downloaded an internet file that looked compatible with my motor and arms. Making some small changes on TinkerCad (a simple-to use CAD software), I printed it out. Turns out, the print did not fit with ANY of my parts.
So, using a caliper, I meticulously measured all of the mounting points and the distance between them to completely redesign the motor mounts. It took a total of 5 iterations (and a lot of waster 3D printer filament) to reach the final design that fit with my brand new motors and carbon fiber arms.
Still, when I printed my final model, the mounts didn't fit onto the carbon fiber arms. For some reason, the prints deformed the circular shape of the mount to an oval. So then I thought. And I thought and thought and thought. How in the world was I supposed to form the prints now? Well, 3D printers heat up filament in order to extrude them. So clearly, the plastic is reactive to high heat. With this knowledge, I heated some water to approximately 170 F and poured it into a bowl. Then, I dipped my deformed motor mounts into the water for 10 seconds. Pulling them out, they were soft to the touch. I was able to mold the mounts around the carbon fiber prop arms for a tight fit. Wow!
Then, it was soldering time. Now I'm no electrical engineer, so I meticulously studied how to solder under the wing of my father. Still, I'm messy, so after much struggle, I soldered my new 40 amp rated ESC's to the power distribution board and mounted these parts across two floors on The Sandwich. I wired the ESC's to the motors and the flight controller, and just like that, it looked like I was making progress!
Then, I had to mount my brand new, 18.5 Volt & 5000 mAh battery. So, I custom-designed and 3D printed a new mount for this behemoth of a battery. After mounting that, it looked like the drone was ready for programming.
But something was missing...
You see, I like to inject a little personality into my every creation. I like to have a personal connection with my inventions. So, I named my drone Frisbee and gave him a face.
Ladies and gentlemen, I'd like you to meet Frisbee!! He's a little shy but I'm sure you guys will get along well.
Week 6: Rest in Peace Frisbee
After a long week of work trying to get Frisbee to fly, it was finally launch day.
And what a tragic day it was. It was March 29th, and the skies were grey. I woke up with a throbbing headache and spilled my orange juice at breakfast. There were still a couple days left for April Fools, but I felt like a fool already.
I made my way to the garage, and there Frisbee was, waiting for his maiden voyage. All of his parts were assembled and ready to fly. The flight computer stood proudly atop The Sandwich, with the power distribution board, ESC's and battery mounted below. The propeller arms were secured (temporarily) with hot glue and zip ties while the motors had the 15 inch propellers attached. Frisbee was basically begging to be flown.
I connected the enormous battery and powered on my radio controller. Frisbee came to life, his buzzers sounding and lights flashing. I put him on wide open ground, and began experimenting with the throttle. To my surprise, everything was spinning perfectly! So then, I increased the throttle for takeoff.
And this is where it all went wrong.
Frisbee lurched up and suddenly rotated on his right side. Before I knew it, he lay upside down with the motors spinning uncontrollably, scratching on the ground. Nuts and bolts flew everywhere, and plastic bits cracked onto the pavement. Frisbee was in major trouble. I disarmed the electronics and rushed him into the emergency room (my garage) to assess the damage.
I was broken, and so was Frisbee. Me, in the figurative sense, and Frisbee in the literal sense. His arms shattered, and the propellers were all scratched up. The flight controller fell off and the GPS module cracked. Oh what a terrible scene.
And so, the rest of my week was spent in the emergency room operating on Frisbee. He didn't make it. I had to do a ground-up rebuild.
I completely redesigned the propeller mounts to be far more durable so that they wouldn't shatter. I tightened all the bolts with Loctite. I designed a new, stronger hold for the flight controller and Raspberry Pi, and I attached the carbon fiber propeller arms with epoxy (which is so strong!). I reprogrammed and recalibrated ALL of the electronics to ensure there were no shorts or glitches. I reconfigured the flight controller so that I could save the drone before another tragedy like this occurred.
Unlike many, I learn from my mistakes. After the devastating loss of Frisbee, I had to change my ways.
From this circumstance, like a phoenix bursting out of the ashes, arose Boomerang. She is Version 2 of my delivery drone.
I made sure that I tested her on grass for the first time. I took care in prepping her and securing all of her parts. And she flew like an eagle. She flew to heights that would've made Frisbee proud. She had some issues, like incessant yawing and sporadic thrust, but these were problems I could manage and hammer out in a few days.
I want to dedicate this post to Frisbee. I still tear up at the thought of him. Rest in peace Frisbee.
Week 7: Propellers and Servos
Welcome back to your weekly installment of delivery drone development! Last week, I got my drone (her name is Boomerang) to hit the skies with pure finesse.
Now with Boomerang ready to fly, I set her to the side and began focusing on other aspects of the drone delivery process. An essential point of innovation on drones is the propeller systems. We're so used to seeing traditional dual blade setups on drones, but there's a whole realm to discover. Drone propellers can affect thrust, RPM, battery life, and sound.
I wanted to specifically focus on sound. You see, I'd hate for a bunch of drones to be flying around my neighborhood creating obnoxious, high-pitched noises while I'm trying to focus on watching The Office. So, I did a deep dive into propeller designs and landed upon the toroidal propeller. It's been used on boats for a while (see Sharrow Marine propellers), but researchers at MIT's Lincoln Laboratories have fitted toroidal propellers onto drones. What they found is that they are far more noise efficient than traditional dual bladed propellers.
I got in touch with Dr. Thomas Sebastian from MIT Lincoln Laboratories who did extensive work with toroidal propellers. Using his research and advice, I 3D printed rough prototypes of the toroidal prop design and attached it to a 920 KV motor to test its noise and thrust efficiency compared to other propellers.
The experimental setup consisted of the following parts:
Digital load cell (1 kg) + HX711 board 3D printed test stand
Arduino Uno
Phone with Decibel Meter app
920 KV Motor
3D printed propellers (pictures displayed in following pages) of equivalent diameter (~10cm)
Propellers: Standard 2 blade, Standard 3 blade, Toroidal 2 blade, Toroidal 3 blade
Electronic Speed Controller
11.1V LiPo battery
Servo tester
Neiko digital handheld tachometer
Laptop to run thrust/load application
To obtain data for each of the props, I set up a spreadsheet to take measurements of RPM, thrust (kg), and decibels . As shown in the images, I initially positioned the phone behind the stand by 15 cm, but later I shifted it 24.5 cm to the left of the stand because the decibel readings fluctuated more due to vortex development in that region. This provided far more precise decibel readings.
I took data for thrust and noise at 1000 RPM increments from 3000 RPM to 10000 RPM. Then, I plotted RPM vs. Thrust to measure lift efficiency and Thrust vs Decibels to measure noise efficiency.
From my tests, it was clear that the toroidal 3 blade propeller was the winner in terms of thrust and noise efficiency. At 10,000 RPM, the toroidal propeller produced 0.18 kilograms of lift as compared to the 3 blade propeller that only produces 0.09 kilograms of lift at the same RPM.
Additionally, when analyzing the emitted sound at 0.04 kilograms of thrust, the toroidal propeller emits the lowest intensity at about 68 decibels compared to the runner-up: the 3 blade standard propeller at 75 decibels.
Because of the cantilevered structure of conventional propellers, they are susceptible to deformations that increase noise through further wind resistance. On the other hand, the toroidal propeller design has a closed loop structure that reduces vibration and dissipates noise. This explains the toroidal 3 blade propeller’s superior performance.
Throughout this test, there are many possible sources of error. Most notably, slight differences in propeller diameter can contribute to errors in thrust and noise efficiency. A larger propeller diameter will increase both the thrust and sound intensity because it is pushing larger volumes of air.
Propeller pitch angle is also an essential variable to consider. On airplanes, there are systems in place to adjust pitch angles. For example, lower pitch angles make the engine reach its maximum RPM at lower lower speeds. On large delivery drones like the ones I plan to build, the variable-pitch adjustment mechanism I developed in the first few weeks of this project may help with range and noise efficiency.
I'm uncertain if I'll be able to mount the toroidal 3-blade propellers onto Boomerang because 3D printed propellers aren't the most reliable. They crack easily, they're uneven, and they're heavy. However, this experiment proves that the toroidal design will be an effective solution for noisy drone propellers. The next challenge to solve would be molding and manufacturing properly designed toroidal propellers for real-world applications.
Week 8: Eyes and Hands
Eyes and hands? After looking at this blog's title, I'm sure you're wondering what that could possibly mean for my delivery drone project.
Well, Boomerang has been capable of flight for a few weeks now, but a delivery drone should be able to do so much more than that. It must be able to hold packages and detect landing spots!
This week I made significant progress in these departments. Starting with the "Eyes", I've developed a tag recognition system that will be essential for Delivery Drone Dropoff Detection (DDDD). The system uses a HuskyLens AI Camera to perform DDDD. The benefits of the HuskyLens is that it can easily be trained to tags like the ones below:
With a press of a button, any of these tags can be saved and trained onto the platform. I've used Python on Raspberry Pi (a small computer) to interface with the HuskyLens so I can get precise data on when the tag has been detected and where in the field of vision of the camera it has been found. This way, down the line, I can send commands via MavLink to the Pixhawk flight controller and autonomously position the drone for precise dropoff.
The landing pad design is yet to be finalized, but it will be unique to each household and it will be made of an array of two tags like the ones above. The larger tag will be used for less precise adjustment, and the smaller tag will be used for precision landing adjustment once the camera's field of vision narrows as the drone gets closer to the ground. I plan to make the landing pad customizable to best suit the user's location. If the user wants dropoffs to be on the grass in their backyard, I will design a landing pad that is green and doesn't stick out like a sore thumb for the end user.
Now, for the "Hands." I've mulled over many designs and ideas for the payload dropoff mechanism on Boomerang. One thought I had was a winch/wind-up mechanism which entails having the package attached to a long string which will be unwound with a motor for dropoff. This design would weigh a lot and also cause great instability during flight when there is wind because the package will jostle around a significant amount. However, this design will be very safe because the propellers can be kept far away from children and pets at dropoff sites. Perhaps I'll engineer this solution down the road.
The design I ended up choosing for the "Hands" was a simple claw grabber I designed with a 9g servo. It will look like this:
My 3D printer has been causing a lot of problems recently, so I've reached out to a friend for my printing needs which has delayed my process slightly. The claws will be mounted under the drone, and I plan to mount three or four claws underneath so that the drone can carry multiple packages. This creates a modular design which will allow the drone to fly to multiple destinations on one trip if it is carrying lightweight packages.
The next step is to assemble the claws and mount them to the drone. Then, I have to program them with the Raspberry Pi. I plan to use an ultrasonic sensor to detect when the drone is approaching the ground so that the computer can actuate the package dropoff mechanism at the right time. Fragile packages should be dropped close to the ground and others can be dropped off from the air.
Well, thanks for checking in again! As usual, I'll see you next week.
Week 9: The Brains of the Operation
As the weeks of Senior Project come to a close, I've begun putting finishing touches on Boomerang to get her ready to present to the masses.
This week, I wired and programmed the brains of the operation, the Raspberry Pi computer.
Boomerang knows to drop a package when an onboard camera detects the smallest of a series of three QR codes detected on the ground. Using the Huskylens camera, I trained the device to recognize tag and received data on the Pi regarding its placement in the frame. Using this data, I am able to determine if it is appropriate to drop a package at a given location.
After confirming the functionality of the camera system, I moved to programming the gripper that holds the packages. Last week, I 3D modeled the system and printed it. This week, with only 10 lines of code, I was able to actuate the gripper so that it opens and closes in a span of five seconds.
Now that I had both the camera and the gripper working separately, I created a new program where I combined their functionality. The logic is simple: if the smallest tag is detected in the frame of the camera, open the gripper! I did run into a few issues in terms of running the programs on the Raspberry Pi as it suddenly stopped connecting to my monitor! I'm used to working with Arduino, so I had to learn how to use SSH (Secure Shell) to control the Raspberry Pi remotely.
Once the systems worked in harmony, I finally mounted them onto Boomerang! The gripper is now located at under the battery and the Huskylens camera is located towards the top of "The Sandwich" near the flight controller.
I've also begun thinking about how I can display all of the technology I've developed as the final presentation date nears. I'm planning to build a stand where viewers can interact with Boomerang using buttons and visual cues.
Over the next few days, I'll start building this display and also take Boomerang for a few test flights with the new tech equipped so that I can fine-tune her flight and ensure that her features are working as intended.
Hopefully, Boomerang survives her next round of testing! Until next week...
Week 10: The Last Mile
The "Last Mile" delivery challenge is something that all companies in the logistics and e-commerce sectors are constantly grappling with. It refers to the final leg of the delivery journey, getting the package from the distribution center to the customer's doorstep. And now, I'm on the last mile of my Senior Project. It's time to get Boomerang from my garage, to the real world!
This means (figuratively) putting makeup on my drone and making her ready for show and display. But before that, I have a series of minor technological issues to hammer out.
Reprogram Raspberry Pi so that it implements while loops instead of if statements to unlatch package. This way, the gripper that holds the package to the drone will stay open while the QR code is detected below the drone, allowing for there to be a larger window for the package to be dropped.
I need to power the Raspberry Pi via the drone's onboard electronics. Right now, it is being plugged into a wall socket. The Pi needs 3 amps current from a 5 volt source and I've already soldered connections from my power distribution board to make way for this load. However, I am currently receiving a UBEC (universal battery eliminator circuit) to step down the voltage from the 18.5 volts of the battery to 5 volts.
The drone needs to be fine tuned for stable flight using the "Autotune" flight mode in QGroundControl. I plan to head to a large, open field to run this and optimize flight.
It would be nice to test the recommended weight capacity of Boomerang. When I'm tuning the flight control, I will bring a water bottle with a carabiner attached to it so that Boomerang can pick it up. I will slowly fill up the water bottle in various flights and weigh it to see at what point is flight significantly hindered by the weight of the water.
Once these issues are hammered out, I'll get Boomerang dressed and ready for her final presentation on May 18th! I've already built a pedestal that allows viewers to interact with Boomerang's various quirks and features. Next, I'll capture video clips of Boomerang in flight and dropping packages so that my audience can see a convincing visual of her in action. I'm also finalizing my presentation and practicing it in front of Boomerang. Trust me, she's a difficult audience to please.
And with that, it's a wrap! It's been a wonderful run. I've learned more than I could've hoped throughout this project. It was a lot more difficult than I anticipated, with leaps and hurdles along the way. This may have hindered the progress I had hoped to make, but it propelled my intellectual exploration and curiosity every step of the way.
Thank you to Dr. Dixit, Ms. Nagpal, and Abishek for supporting me and keeping me on track to get Boomerang in the skies.