Autonomous, Wirelessly Controlled Hovercraft

This is an instructable that takes you through an entire month and a half of work, designing and creating a wirelessly controlled modular hovercraft, that can be controlled with an RC controller, or made completely autonomous. I'll walk you through how we built our hovercraft, including all circuitry, firmware, and software. This instructable will also include ways to build a significantly cheaper hovercraft with all of the same functionality, but slightly lower performance. Lastly, I'll show you some of the things we learned, as well as some propositions for "Big Bird 2.0." Hope you enjoy it!

-Bradley Powers
bpowers.org

Step 1: Chassis Fabrication Using Blue Insulation Foam and Carbon Fiber Hand Layup.

In this step, I'll go over how we made the chassis for our hovercraft, as well as how you could make one without all of the mess. I don't have any pictures for this step, as I was literally covered in blue foam chips or epoxy resin.

The chassis was made using blue insulation foam, which can be purchased at home improvement stores, coated in carbon fiber composite. The blue foam was cut to specification on a CNC mill (Cad file will be included, picture shown here), and then coated in composite using a process using hand layup. I don’t have any pictures of the process, as I was covered in sticky gross resin, but I’ll try to explain (this is a very easy process, it is important to take your time and make everything look good). To begin, you want to get all of your materials and supplies for the hand layup. MAKE SURE YOU ARE WEARING NITRILE gloves, latex won't cut it with epoxy resin!!!

To begin, it's a good idea to just do a dry run. We basically cut all of our Carbon Fiber fabric (5.7 Oz/Sq Yd, 50" Wide, .012" Thick, 12.5 x 12.5 Plain Weave, available at [http://www.fibreglast.com/showproducts-category-Carbon%20Fiber%20(Graphite)%20Fabrics%20&%20Tapes-15.html Fibreglast]. We cut a piece for the top and the bottom, leaving about 1" of overlap on each side. We also very carefully cut holes for the duct, and for the "pocket in our design. We also cut strips for the inside of the "Pocket" as well as for the sides of the chassis. Then we cut reinforcement strips, basically so we could get extra stiffness where needed. Lastly, we cut a strip to line the inside of the duct, which ended up working quite nicely, as we were able to cut holes for the holes in the duct (which, by the way are CRUCIAL) which feed air into the skirt.

Now for the fun part! No joke, put on your nitrile gloves, and clothes you don't like. One other very important thing to do is to find a well ventilated area to work in, or to wear a ventilator. We used a chemical lab fume hood. Now that safety is covered, we can get into the fun stuff. The very first thing that you need to do is mix your resin. We used System 2000 epoxy resin, with 2060 epoxy hardener, available here. We used about one pint of resin, and about a third of a pint of hardener, which you mix in a 3:1 ratio by volume. It is very important to stir that very well, as it will ensure that all of your epoxy actually hardens. Next, paint that mixture on the bottom of the chassis (for example) very generously, and then place your CF fabric on the chassis where you want it. Then, press down, and watch the resin soak through the fabric. You can use a squeegee to make sure that the resin wets out the carbon fiber everywhere. At this point you can either let it dry, or cover it in tinfoil and move on. Basically from here, rinse and repeat. Keep in mind that the better you do the hand layup, the less grinding out rough spots you have to do. Take your time, and make sure that everything is as you want it, as working with carbon fiber when it is dry is basically no fun.

There are a few other ways to go about making the chassis. You could completely skip the carbon fiber process, and just use foam, but you will probably want to go with an EPP foam, as it will stand up to abuse much more than other foams. You could also fabricate it out of wood, or use sheet metal bent to shape. Really, all that matters is that you can fasten things to your chassis, and that it has the proper holes to make the lift fan and skirt work.

Step 2: Lift Motor System

For lift, we used a Team Novak 10.5 turn Brushless motor, and a Vario Pitch size 6A propeller with 4 6” blades and a 3.2 mm hub (available here). This worked really well for us, as the brushless motor provided significantly more power than we actually needed, so we could carry weight on the order of 5-7 pounds. The variable pitch propeller allowed us to adjust the pitch so that we could find a good setting that allowed for maximum runtime. A significantly cheaper version of this lift system would be a standard speed 540 motor with a generic 6 x 4 pusher prop. This would not allow as much weight (bad, might not matter) on the craft itself, but is a good system that can be purchased for very little (good).

Step 3: The Skirt

Ahhh, the skirt. We used ripstop nylon which can be purchased at a fabric store, and made ourselves templates (courtesy of Irgsmirx, from rc-hovercrafts.com) which we used to cut the fabric. We then used double sided tape (not the foam kind, this stuff is literally just adhesive) to connect the pieces of the skirt together, as well as to connect it to the chassis. We did this mainly because it was vastly easier than sewing, and worked fairly well. It is very important that the skirt is smaller on the bottom than the top, it pulls the skirt toward the underside of the craft, which helps maintain a good air cushion (Picture 2).

Step 4: Propulsion

For propulsion, we used two E-flite Rimfire brushless motors with V-Pitch props, which allowed us to move forward, backward, and pivot, all without varying motor speed, which means that response time was extremely quick. Again, this could be accomplished with a generic brushed motor and prop, which would increase response time on controls (bad), but would also significantly lower cost (good).

Step 5: Parts List

Lift system:
Team novak 10.5 brushless motor and GTB speed controller
Ripstop nylon cut to template

Propulsion system:
Rimfire 22-1000 Brushless motor with V-Pitch propellor system
E-Flite BL-12 Brushless motor controller
Custom Cut 6061 aluminum motor mounts

Chassis
CNC milled blue foam core
Carbon Fiber hand layup shell

Electronics
PIC 18F2455
Other electronics

Step 6: Electronics

Circuit Overview

Project Big Bird required two circuits, each with specialized functions:

Onboard PIC circuit:
send/receive data from host PIC via wireless
controls the lift fan and propulsion fans via PWM output signals
reads sensors (convert analog to digital)
Host PIC circuit:
send/receive data from host computer via USB
send/receive data with host PIC via wireless

Photos of both are shown below.

Circuit Diagrams also appear below

Power
Onboard:
Takes in +6V from Novak Electronic Speed Control (ESC) Power Output
Runs entirely on +3.3V, produced by a voltage regulator

Host:
Runs on +5V from host computer
XBee module uses +3.3V, produced by a voltage regulator

Microcontrollers (PICs)

We choose to utilize microcontrollers from the Microchip Technology’s PIC18F family due to availability and our previous experience. The data sheet for this class of microcontrollers can be found here.

Electronically, these PICs require between 2.5-5.5 V, well within the range of both circuits.

Wireless Module

To achieve wireless data transfer between the host computer and the onboard lift and propulsion fans, we chose to use a XBee Pro OEM RF Module, as shown below

The XBee provides a low cost, low power solution to the wireless communication. It offers a 300 ft range over a standard 802.11 frequency. The device is ready to operate right out of the box with minimal support circuitry. A pin diagram is shown below

Pins 1 and 10 must be connected to +3.3V and GND, respectively. Pin 2, the XBee’s data output node, connects to pin 18 of each PIC. Similarly, pin 3, the XBee data input, hooks in to PIC pin 17. These four connections are all that is necessary to send and receive byte data wirelessly between PICs.

For more information on the XBee OED RF module, check out the manufacturer data sheet.

Lift Fan Electronics

Big Bird’s lift fan is a Novak GTB/SS 10.5 Pro Brushless System, which includes the motor, the speed control, and other necessary components.

This system requires 4.8-7.2 V to run, which motor performance increasing as the input voltage approaches 7.2 V.

The lift fan system’s electronic speed control (ESC) provides a +6V power output alongside the white signal cable which provides PWM control of the lift fan. This +6V was down-regulated to +3.3V and used to power the onboard circuit.

We chose to power Big Bird’s lift fan system with a 7.2V 2-cell Team Orion Carbon Edition Lithium Polymer battery.

The choice of Lithium Polymer for the battery type allows considerably longer run-time and lighter weight than other options, such as NiMH (Nickel Metal Hydride). The power density of these batteries, as well as their availability are why we went with these batteries.

For more information on the Novak GTB Motor control system, check out this link.

The Team Orion battery specifications can be found here.

Propeller Electronics

To provide forward and reverse thrust to Big Bird, we chose to use 2 Electri-Fly Rimfire 22M-1000 brushless motor systems.

These motors offer variable pitch-control of the propeller blades, allowing us to quickly shift between forward, reverse, turning, and at-rest behaviors all at a constant motor speed. Variable pitch control avoids excessive wear and tear on the motor’s mechanical components and also prevents the propulsion battery and motor electronics from experiencing too much variability in voltage and current loads, which are clearly ideal for prolonged life.

The Rimfire motor specifications indicate that the device will operate between 7.2 and 12 V, though our experience showed that the effective range was more like 9-12 V.

The specifications for the Rimfire motor can be found here.

To power the Rimfire motor optimally (as close to 12 V as possible to achieve the greatest potential thrust), we chose to utilize a 11.1 V 3-cell Li-Po battery made by Hi-Model, shown below

More information on this battery can be found here.

Ultrasonic Input

To provide a proof of concept sensor for the Big Bird platform, we chose to add an ultrasonic distance sensor to the final prototype. Based on low cost, availability, and plug-and-play ease of use, we elected to use the Maxbotix EZ1 ultrasonic sensor, shown below

The EZ1 requires 2.5 – 5.5 V, and for our implementation was powered using the 3.3V power of the onboard circuit. This sensor’s specifications indicate that it offers ~6.4 mV/in analog output at this voltage, a factor which we found to agree well with observed values.

The pin diagram of this sensor is shown below

We connect Pin 1 to GND and Pin 2 to the +3.3V of the onboard circuit.

The sensor’s Pin 5 (AN) provides analog voltage output scaled to the distance reading. We connect this pin to AN0 on the onboard PIC, where it is converted into a digital reading and sent wirelessly to the host computer for display.

The datasheet for this ultrasonic sensor is found here.

While the sensor currently has no integral part in the performance of Big Bird, we believe it offers a nice proof of concept for Big Bird’s sensor potential. With 5 analog inputs (AN0 – AN4) available on the onboard PIC, plenty of future options for navigation control exist with the proper code modifications.

Step 7: Firmware and Software

Wireless control
We decided that having wireless control of the hovercraft was an absolutely essential feature, so we acquired two XBEE wireless modules. We set up two circuits, the host circuit and the onboard circuit. Using the PICs' ability to send and receive signals, we set up the host PIC to send a pulse width specification to the onboard PIC. Once we were able to send signals from the host PIC to the onboard PIC, we added the XBEE modules. The XBEE modules required no additional code modifications; we could send and receive signals as if the PICs were connected by invisible wires.
Receiving signals
We wanted to be able to receive data from sensors mounted on the hovercraft. We decided to use PORTA for analog sensor inputs and PORTB for digital pulse width outputs. This decision was motivated by the fact that only PORTA ports could act as analog inputs. We set up five pulse width outputs and controlled them with five sliders on the computer. Once this was working, we added an analog input that could send its signal from the onboard PIC to the host PIC and finally to the computer. When we added additional analog inputs, however, we ran into problems. The first problem was that it simply took too long for the PIC to read from all of the inputs and send the signals to the host PIC. The delay was long enough to affect the pulse width outputs. We fixed this problem by limiting how often the PIC would read from the analog inputs. The second problem was that the analog inputs seemed to be reading and sending the wrong signals. We found out that we needed to wait longer in between reading different analog inputs. After implementing this delay, we still struggled to effectively read multiple analog inputs. We settled for the ability to only read from one sensor at a time.
For sensors, we experimented with an accelerometer and a sonar sensor. We found that we could sense tilt with an accelerometer, but we could not sense enough tilt for the accelerometer to be useful for our purposes. We found that the sonar sensor was effective for measuring distances greater than six inches, and was accurate to within about three inches. We calibrated the sonar sensor and set up the computer to output its reading in inches.
User interface
Up until the final week, the user interface consisted of five sliders, one for each pulse width output. In the final week, we thought more about how the user might want to control the hovercraft. We settled for a slider to control the speed of the lift fan, sliders to control the speed of each propulsion fan, and sliders to control the pitch of each propulsion fan. For the user, controlling all of these sliders was a difficult experience, so we added key combinations. G and H controlled the speed of the lift fan, 1 and 2 controlled the speed of the left propulsion fan, and 9 and 0 controlled the speed of the right propulsion fan. Up and down moved the pitch controls in the same direction (for backing up and moving forward), and left and right moved the pitch controls in opposite directions (for steering).


Here's how you can get this up and running yourself. First of all, you need libusb installed on your computer, so that the host PIC and your computer can communicate. I won't go into how to install libusb, as it varies between different operating systems. Also, you'll need a way to flash the PICs with the firmware provided. Microchip's Pickit 2 is what we used, along with their MPLAB software.

Step 8: Putting It All Together

To put everything together, we used custom motor mounts that we had cut out of 6061 aluminum on a water jet to our specifications. We then bent these to spec using a hand brake. These motor mounts were fastened to the chassis using Click Bond Fasteners, which are specially designed for use with composites. We used these because it is very hard to work with composites after they cure, and these fasteners merely bond to the carbon fiber using an epoxy. We mounted all circuitry to the chassis using double sided tape.

Step 9: Next Steps

So, now that a hovercraft is built, where do we go from here? There are quite a few next steps that our group was interested in pursuing, but didn't have time. The first big idea we had was working on making the hovercraft autonomous. There are a number of ways to go about this, but they're a little outside the scope of this Instructable, so I'll leave it for later. Another fun thing to do, is hook everything up to a Radio Controller. This basically gives you a radio control hovercraft.

Thank you for reading, I hope you have enjoyed this. If you have any questions, we have a website:
www.bradleyp.com/BigBird

Also, feel free to email me at [e-mail], I'm more than happy to help with designs, make recommendations, or listen to your recommendations for future research.

Thanks again.