CPE-322 Homework 5

Autonomous Drone Delivery

Prof. Hong Man

Due: 3/4/14

“I pledge my honor that I have abided by the Stevens Honor Code”

Group Members:

Michael Paulauski______

Eddie Bowlby______

Matt Leslie______

ShaQuill Thomas______

Section 1

Summary of Group Work Assignments

Michael Paulauski will be covering the quadcopter portion of the project. He will be discussing the various implementations of quadcopter design, as well as manufacturing and assembly of quadcopters via 3D printing.

Edward Bowlby will be covering the computer vision portion of the project. He will be discussing existing visual recognition advances and technologies as well as how they could be applied to an autonomous delivery drone.

Matt Leslie will be covering the GPS and tracking portion of the project. He will be discussing how GPS tracking can be applied to this type of drone project, as well as how parcel tracking is performed in modern shipping.

ShaQuill Thomas will be covering the charging station hub portion of the project. He will be discussing the implementation of the charging station that will be used to charge the autonomous drone system throughout the day in order to minimize downtime and latency between deliveries.

Michael Paulauski / Eddie Bowlby / Matt Leslie / ShaQuill Thomas
Percentage of effort towards this assignment / 25% / 25% / 25% / 25%

Section 2

Quadcopters

Quadcopters come in many different shapes, sizes, and designs. They are some of the most popular devices currently being produced by 3D printers. Thy can vary from extremely thin, spindly, and stark designs, which are only capable of lifting themselves, to complex and intricate designs, able to lift several pounds of cargo in addition to themselves. The design and manufacturing of a quadcopter fundamentally dictates its capabilities, durability, weight, and power. Traditionally, quadcopters are unique from other helicopter designs in that instead of one main lifting rotor and one counter-balance rotor, they have 4 rotors arranged radially from a central hub.

A very sleek and minimalistic quadcopter.

However, besides this fundamental four-rotor principle, there is very little that cannot be changed about the quadcopter design. As most quadcopters are at least partially 3D printed, careful consideration must be made as to the density of the plastic pieces that make up the quadcopter’s body. If these pieces are too dense, it will negatively affect the power:weight ratio, the flight time, and the maneuverability of the quadcopter. However, if they aren’t dense enough, they could easily crack and splinter during a rough landing. Since our design will require both wireless communication as well as onboard computing and vision processing, we need to consider the weight of these components when summing up the total weight of the quadcopter + the package that it is carrying.

However, scientists have recently discovered a possible new design for quadcopters that will allow them to lift themselves far more efficiently, gaining up to 25% more flight time. This new design has one central lifting rotor and 3 radially organized control motors that are angled at 45 degrees. This design allows for not only longer flight time, but also greater maneuverability and lifting power.


Computer Vision

The implementation of the computer vision system is subject to many physical, economical, and ethical constraints that are important to consider during the design of the system. Steven Zucker of McGill University in Canada divides the physical constraints of computer vision systems by comparing them to the physical constraints of human vision. He divides the physical constraints into “three main categories: computational, behavioural, and implementational.” He explains that the computational constraints of systems involve the mathematical ability to transform the view of a three-dimensional scene into two dimensions. Computational constraints are essentially the software engineering challenges that being researched internationally. Behavioural constraints are the limits on the reactions the vehicle can have after the “vision” of the environment has been processed. Lastly, the implementational constraints are the physical restrictions, in the scientific sense of the word, that determine what goals the vision system can be implemented to achieve. Surprisingly, developers have continued to extend the boundaries on all of these physical constraints of computer vision systems, as evident by multiple implementation successes from Google’s autonomous car to Amazon’s delivery drones. With respect to our project, the implementational constraints are certainly the biggest challenge. It will take significant planning and possibly innovation to implement all of the hardware and software that are currently necessary for a robust computer vision system in an autonomous vehicle that is planned to be significantly smaller than comparable existing autonomous vehicles.
The second group of constraints that significantly impact the design and implementation of the computer vision system are economical restrictions. If designing a functioning computer vision system wasn’t challenging enough, it must also be accomplished in an economically feasible way. The end goal of our project would be replace the need to stop package delivery trucks and get out of the vehicle to deliver packages, instead utilizing an army of autonomous drones to deliver packages from a moving truck to the appropriate homes. This introduces the necessity of the project to make “business sense”- the increase in package delivering efficiency and other cost savings must more than offset the cost of development, implementation, and mass production of the drones. The development and implementation of the computer vision system arguably is the most expensive aspect of the project. According to a Washington Post article, a somewhat comparable autonomous vehicle, the Google autonomous car, currently requires $70,000 in 3-D sensors alone. It will be important for our group to research precisely how much more efficient drones would make the package delivery business and to what extent costs could be cut and operations improved with their implementation, as the total economic benefit a drone provides restricts how much can be spent on the drone’s development and implementation.

Ethical constraints make up the last, and arguably the most important, group of design and implementation restrictions for our project. In the design and implementation process, we as engineers have an ethical responsibility to develop and implement the drone to the best of our ability and with safety as the highest priority. The computer vision system, which will largely govern how the drone to moves and interacts with the environment, especially must be developed with safety in mind. The drone must have multiple failsafes that ensure that the drone never causes harm to any person or object or creates a situation where harm could come to a person or other object. Additionally, our group must be incredibly attentive to legal constraints regarding autonomous vehicles and robots. Autonomous machines are in their youth, and thus the legal system has not yet caught up and decided who is responsible for accidents involving autonomous machines.

GPS and Tracking

As mentioned in our previous report, a crucial component of any unmanned aerial vehicle or drone system is being able to track the positions of any number of drones at any particular time. A camera system located on the drone itself can help with navigation, but in this consumer application more is required on both the shipping and consumer end. Similar systems for locating an item or vehicle are already in place for things like modern smartphones or navigation tool for automobiles. Also, as previously mentioned in an earlier report, this is also popular in the airline industry.

For our project and our drones, we would need some form of small scale asset tracking system to track our drones. Asset tracking is a system that can have a GPS module track the location of the drone coupled with an “active” RFID, in which a radio or cellular network transmitter then transmits the location data a long distance, theoretically to some form of center when then interprets the data in a way that it can be seen by the end use and used to actually located the asset.

As for the device that will actually be doing the device tracking, our project requires some form of small and lightweight GPS tracking unit. The most common type of GPS tracking unit, as well as the type best suited for our project is the “data pusher” type. In addition to receiving GPS location data from a satellite, this type of tracking unit pushes the data to a determined server. Unlike the “Data logger” type of tracking unit, where the data is stored on the unit itself, this data pusher unit has all of the location data analyzed on stored on the server. There is a large selection of software suites on both traditional PC and mobile platforms to interpret this data and present it to the end user. For our project, as mentioned in our previous report, I feel it would be easiest to develop or use pre-existing software. This would allow us to have the data transmitted over a GSM mobile network. Obviously this infrastructure is already in place, and would make it very easy for us to receive data from our tracking unit.

As for individual package tracking, UPS uses a system that involves a UPC code being placed on each package they carry. Each time a unit leaves or arrives at a UPS facility, that UPC is scanned. This allows the employees that scanned the package to see where it came from, and where it is going. Because their scanners are also connected, this information is then sent to a server where it is shown on UPS’ website. Each UPC also has a tracking number attached to it, which users can search via the website or a mobile application. The tracking numbers pulls all scans associated with the package, and present them to the user. It is this system that allows users to see when exactly their package enters or leaves any UPS facility. This also allows them to see at what step and approximate location their package was lost, if such an event were to occur.

Charging Station Hub

The implementation of the charging station hub that would be required in order to charge the drone throughout the day for delivery, consists of a simple design of magnetic and inductive charging principles already established in today’s current technology. From a technology article published by the Society of Automotive Engineers International, it has been looked into that a team developed a wireless, contact free power transfer mechanism that is safer and robust to imperfect alignment on landing at the base station and that void back to the launch site for recharging off power lines for a drone system. The development of this research goes on into noting that a magnetic field is created using inductor coils on both the transmitting and receiving sides. By using small induction coils around the drone to increase efficiency and decrease interference. This will lead to faster recharging compared to just using one coil. Once charging is completed, the drone would send a signal to the base station to open the drone to fly away. This development can be very useful for our drone delivery project in which we can analysis the foundation that has been created by this publishing to instrument a working charging station to be used in our project.

After further investigating the possibilities of magnetic and Inductive charging method for the drone system, a video on Youtube showed a simplistic model to create the charging station hub. The video provided a visual depiction on how the charging functions with the drone, and how we can keep track the position of the drone using computer software. The video goes a little more in depth on the types of tools and equipment used for the design. Some of the equipment include, a computer, charging unit, voltmeter, motion sensors and a force sensitive resistor (Pressure Sensor). The video showed trials of a drone system detecting a charging station hub and then positioning itself to land on the station which is calculated for accuracy on a computer software program. Screenshots from the video for the drone detecting the charging station process is shown in Figures 1 & 2 below.

The last resource found when researching this aspect of the project was from two articles, one regarding “The future: Quadcopter UAVs recharging your smartphone with wireless power” and “Quadcopter Automatic Landing on a Docking Station”. These two article bring to our project different possibilities that can further improve the way our drone system can function and also enhance the way the charging station interacts with the drone. The first article proposes that the drone system would be able to wirelessly transmit power to nearby devices. Which is a possible method in which if properly implemented can cut half the required charging station needed for delivery. The second article provided a further in-depth analysis on the computer vision algorithm needed to properly have the drone system land on the charging station. Picture of the articles theory on positioning is found in figure 3. Figure 1 Drone Approaching Charging Station Hub Figure 2 Drone Aligning itself to Charging Station Hub

Figure 3 Picture of the Drone System collecting information needed before landing

Section 3

1. Identify realistic constraints of your design. In particular, you should focus on the following constraints, a) economic, b) environmental, c) health and safety, d) manufacturability, and e) sustainability.

Some economic constraints that our design could run into is the fact that delivery companies might not want to invest in re-designing their fleet of trucks to support our drone deployment platform, as it might be too high of an up-front cost. The environmental constraints would mostly be surrounding the construction and assembly of the drones, which would be made out of high-impact plastic. The health and safety constraints would mostly revolve around the drone’s ability to use computer vision to guide itself around obstacles, including people and animals. Manufacturing constraints include the need for interchangeable parts for easy replacement, as well as quick manufacturing times for a constant built-out of a new fleet. Finally, sustainability constraints must be considered, such as a nearly-constant manufacturing time for replacement parts, as well as the ability to recycle old, worn-out parts.

2. Identify possible professional and ethical responsibilities of your system or component

or process.

We have a responsibility not only to deliver packages safely and quickly, but to make sure that the delivery of said packages doesn’t harm anyone else, whether by direct impact, or by unsustainable business practices.

Section 4

Based on your research and study on the topic of your project, itemize all project objective attributes, and construct an objective tree with constraints. Refer to Lecture 4 for examples. The objective attribute list should be as complete as you can achieve, and each item should be briefly explained (in one or two sentences). Your first construct of the objective tree may not be entirely correct. You will have chance to modify and improve it in future steps.

Attributes

●Safety

●Economic

●Travel at safe altitudes

●To Prevent Collision

●Manufacturing

●Re-designing of delivery trucks

●Assembly of the drones

●Interchangeable parts

●Addition of charging station hubs

●Addition of deployment platforms for packages

●Computer Vision Algorithm

●Ensuring Accuracy

●Maximizing Deployment Time

●Minimize Delay

●Nearly-constant time for replacement parts

●Ability to recycle old, worn-out parts

References

Regarding Quadcopters

Regarding GPS and Tracking:

Regarding Computer Vision

Steven Zucker, Computer Vision and Human Perception



Regarding Charging Station Hub

Article: Inductive or Magnetic Recharging for Small UAVs

Video: Inductive or Magnetic Recharging for Small UAVs

Article: Quadcopter Automatic Landing on a Docking Station

Article: Quadcopter UAVs recharging your smartphone with wireless power

your-smartphone-with-wireless-power