Back to Roger's Design Portfolio

Faculty Representative for the
United States Naval Academy's
Autonomous Underwater Vehicle Club

This project and page are both works in progress.

AUV mockup in tow tank
AUV mockup in tow tank for drag testing.

Tow Tank Testing Movie (Quicktime 5 MB)

AUV Configuration and Programming Guide (PDF 150 KB).
Work in Progress


  1. Project Background
  2. Vehicle Functional Requirements
  3. Configuration Ideas
  4. CPU Ideas
  5. Hull Ideas... The Tyranny of the PVC Pipe Idea
  6. CPU Administration
  7. Control Surfaces
  8. Depth Cell

Project Background

Being the faculty representative for a project like this is tricky. You want the students to do as much of the work as they can and at the same time you want to steer them away from the really bad ideas. Additionally, in a project as complicated as this, there will be some aspects which are just too hard for the students to do in the time they have allotted for the project.

This club was formed with the intention of building a vehicle to compete in AUVSI's Student AUV Contest.

Before the student's got involved, the other faculty representative and I felt it necessary to select and become familiar with most of the major components for the vehicle. This allowed us to provide some good instruction for the students on interfacing with the hardware. We selected the following components:

Vehicle Functional Requirements

One of the first steps that we helped the students work through was determining the functional requirements (FR) of the vehicle. These FR's were determined from two sources: the rules of the AUVSI Contest and other envisioned uses for the vehicle.

    Contest Imposed or Derived FR's

  1. Minimum Weight: We set a goal of less than 40lbs (There is a point penalty for heaver vehicles).
  2. Able to drop a 1/2 lb weight at a particular location
  3. Able to home in on and touch a light source with particular characteristics.
  4. Able to follow a PVC pipe on the bottom of the pool
  5. Maximum depth 25 feet.
  6. Perform 3 contest attempts in 15 minutes
  7. Swim straight through a validation gate
  8. External kill switch which cuts all power to the vehicle.
  9. 5 percent positively buoyant.

    Additional FR's

  10. Able to wirelessly communicate with and reprogram the vehicle.
  11. Able to add payload modules for future student projects.

Configuration Ideas

We had the students think of many different ways that thrusters and control surfaces could be arranged to maneuver a vehicle in three dimensions. Here is an example of three configuration considered.

Thruster Configuration 1

early configuration 1
Port and starboard main propulsion thrusters. Fore and aft vertical thrusters for depth and pitch control.

This configuration is one of the more common configurations seen at the AUVSI competition. The only direction this vehicle cannot travel is sideways.

  • AUV can hover, yaw, and pitch in place
  • Control algorithms are the simplest
  • Heaviest and most costly configuration

Thruster Configuration 2

early configuration 2
Three main propulsion thrusters arranged in a triangle (when viewed from stern).

This design uses only three thrusters, thereby saving weight and money. All three thrusters are main propulsion thrusters. By operating the thrusters in a coordinated fashion they can control both pitch and yaw. Stationary depth holding is more difficult. All the thrusters can be used for forward propulsion, so this is a more efficient use of thruster mass. By comparison, configuration 1 only uses 50% of its thruster mass for forward propulsion.

  • Less weight and cost compared to the 4 thruster AUV.
  • All thrusters can be used for forward propulsion, therefore higher vehicle speed or smaller thrusters are possible.
  • AUV must be in motion to control depth (via pitch).
  • Control algorithm is more complicated to coordinate pitch and yaw between 3 thrusters.

Thruster Configuration 3

early configuration 3
This configuration uses a single main thruster for propulsion. Control surfaces are used for pitch and heading control.

This configuration only uses a single main thruster for propulsion. Pitch and heading are controlled by control surfaces.
  • Likely to have the lowest power consumption because there is only 1 thruster.
  • Likely to be the smallest possible vehicle.
  • Vehicle must be in motion to control depth, heading, and pitch.
  • Difficult to hover
  • Control surface linkages make inefficient use of interior volume and provide more leakage paths.

Other Thruster Configurations

Many other thruster configurations were considered. Including:

Here is a configuration that I came up with. It uses 8 bilge pump to allow an AUV to travel: forwards, backwards, pitch up/down, yaw left/right, slide sideways, and move vertically. The piping is arranged such that the pump suctions aid the direction of intended movement in most cases. Too bad bilge pumps are a very inconsistent source of thrust.

* bildge pump thruster configuration
A Solid Works sketch model I made of my bilge pump maneuvering unit.

Figuring out which pumps should run to move the vehicle in each of the above mentioned directions is left as an exercise for the reader.

Configuration Evaluation

Then I took the students through the exercise of selecting criterion for comparing each concept. Below are the results from one round of comparisons.

early config ranking
The rankings for configurations 1, 2, and 3. This round of comparisons indicate that configuration 3 is preferable to configurations 1 or 2. See the text for why we chose to NOT use configuration 3.

The above figure shows a set of comparisons which the students made. This comparisons indicate that configuration 3 is preferable to configurations 1 or 2. There we a couple of problems with this comparison.

AUV mockup in tow tank
The final configuration for the first version of the vehicle.

The final thruster configuration for the first version of the vehicle is not one of the configurations listed above. In an attempt to save weight we decided to use a port and starboard main thruster and control surfaces to control depth. Personnel usage was another factor in deciding to try to build a set of control surfaces. During the semester there were 12 students working on the project. Designing and building the control surfaces and their actuators in a waterproof housing was a full time project for three of them.

The control surface project was a chance for those students to take a project from the initial concept to the final detailed design through to the construction. This was a fantastic learning experience which is the point of the exercise of course.

As it turns out, building a waterproof box that is lighter than a pair of small commercially available thrusters is harder than they thought. It is likely that the control surfaces will be replaced with a pair of thrusters for depth control.

CPU Ideas

The table below lists the components we wanted the CPU to interface with:

EquipmentNumber of Interfaces
Radio Modem1
Tethered Comms Link1*1*
4 Thrusters44
Depth Sensor1
2 Cameras2*2*
Marker Release Mech1*1*

A number of factors went into selecting the CPU for the vehicle. These included:

Technologic Systems, TS-7200 ARM Based PC104 SBC

The TS-7200 Single Board Computer

  • Cheap (approx $200 each)
  • Low Power (less than 1 Watt)
  • PC104 Bus Allows Expansion
  • 8 Channels of A/D, 2 USB Ports, 2 Serial Ports, and ethernet on base board.
  • Full Debian Linux File System on board (1-2 Gig CF Card).
  • Onboard C compiler
  • Complete console via a serial port (therefore can be used via a radio modem).
  • Onboard ethernet allows multiple high speed connections.
  • Requires an understanding of Linux system administration
  • Power up time is approximately 15-30 seconds. Longer after improper shutdown.*
  • File system may become corrupted on loss of power if vehicle is not shutdown properly.*
Rabbit 3000 Microprocessor

The TS-7200 Single Board Computer

  • Students use the rabbit in much of their other course work, so they are familiar with it.
  • No significant boot time
  • Handles power cycling well
  • Very low power
  • Very small
  • Very cheap
  • No way to reprogram wirelessly
  • All wireless communication functions would have to be custom written.
  • Not powerful enough to do the required image processing.
  • Expansion only possible via serial port (slow!!)
  • Very limited program and data memory.
Pentium Based PC104 SBC Arcom and Parvus are two big manufactures of single board computers.

The TS-7200 Single Board Computer

  • PC104 bus allows expansion
  • Powerful enough for image and sonar processing
  • Possibility of using other languages besides C (Java, Python)
  • Much more expensive $500+ per board
  • Highest power consumption and heat generation
  • Fewer ports on the base board than the TS-7200, therefore requiring more expansion boards.

*Running the computer from onboard flash drastically speeds up boot times (less than 5 seconds) and provides a journaling file system (which handles inadvertent loss of power much more gracefully). But the gcc compiler and development environment will not fit on the onboard flash so compiling must be done off the board.

I selected Technologic Systems TS-7200 for the AUV CPU. I continue to be very impressed by this board. It is an excellent choice for projects like this.

Hull Ideas... The Tyranny of the PVC Pipe

"Lets just get a piece of PVC pipe and make the hull out of that!"

Anyone who has every tried to build an AUV for the first time

CFD Sketch of a round tubed vehicle
The only benefit of the PVC pipe hull is hydrodynamic.

Here are the problems with a PVC pipe hull:

The above reasons are why I selected Pelican Cases to provide the waterproof enclosures for the vehicle. They provide:

CPU Administration

I have started writing an AUV Programming Guide (PDF). Currently, it covers the initial configuration of our ARM based single board computer. This too is a work in progress.

Control Surfaces

We decided to try to build control surfaces to control the vehicle's pitch and depth. For this to be worth while the control surfaces and their actuators should weigh less than a pair of thrusters. Otherwise it is simpler to just use a pair of thrusters for depth control. Additionally, this was undertaken to give three of the students a small self contained project which would take them all the way through the design process.

"Its a waterproof box with a motor inside and shaft coming out.
How hard can that be?"

A student to remain unnamed

The basic vision was a waterproof box with the shaft for the control surfaces coming out either side. This "servo box" could then be mounted on the AUV's frame in the desired location.

Different types of actuators were considered, including stepper motors and hobby servos. The students selected hobby servos as we already had circuit cards to control them via serial commands.

Once servos were selected, the students considered several mechanisms to connect the servo to the shaft. These included linkages, gears and hydraulic rams. The students elected to use a linkage.

An early sketch model of the servo box
An early sketch model of the servo box.

An early sketch of the proposed shaft seals
An early sketch of the proposed shaft seals.

The final servo box was a folded and welded sheet metal box. The shaft seals were tapered cones and o-rings. The there was a lot of wasted space in the box. Therefore, weight had to be added to the vehicle to make it neutrally buoyant. This causes the servo box to weigh much more than two thrusters. Additionally the linkage connecting the servo to the shaft is too flimsy.

Here are some photos of the final servo box for the control surfaces:

Final Servo Box Drawing Servo Box Photo 1
Servo Box Photo 2 Servo Box Photo 3

Depth Cell

Initially the students tried to use a depth cell from the past vehicle. This cell required an external bridge circuit and regulated power supply. Additionally, since the cell was designed to be mounted in a pipe the end with the wires was not waterproof. So it had to be potted in epoxy, since the depth cell on our vehicle would be mounted on the frame with an external cable running back to the CPU housing. Calibrating the bridge circuit also required making a 30 foot water column.

Depth Cell Calibration 1 Depth Cell Calibration 1

Calibrating the old depth cell.

This old depth cell was replaced by a fully waterproof cell which can take 13-28 VDC, and provide a linear 0-10 VDC proportional to 0-30 feet of depth in a single compact package requiring no external circuitry.

The new completely self contained depth cell
The new completely self contained depth cell

This project and page are both works in progress.