PSU 541: Artificial Intelligence
Professor: Bart Massey
Fall 2004


By David Allen, 11-15-2004


Project Prospectus: PSAS Rocket Simulator


Introduction:

The Portland State Aerospace Society (PSAS) [1] has the goal to be
the first amateur group to build an actively guided rocket. The group
needs a flexible simulated environment to test current and future
intelligent avionics systems.

I will rebuild and expand the current simulator to create a modular
engineering tool for testing the fully integrated rocket, individual
systems, and prototype avionics. For the first version of the
simulator, I will only implement relatively simple models of the
environment, rocket body and rocket systems; but I will build the code
to allow easy expansion and replacement of these first models. The
code will be open-source and I will document the work on the PSAS Wiki
site [2].


Current Simulator and Needs:

The current simulator [3] was written by Bart Massey in Nickle [4]. It
is a basic simulator for a rocket with a single boost stage, a drogue
chute and a main chute. It simulates a purely vertical flight path
with acceleration applied by rocket boost, by gravity and by friction.
The state of the rocket is used to generate system specific messages
that contain pressure measurements, inertial measurements, and launch
pad umbilical disconnects. Commands to ignite the engine and deploy the
parachutes must be manually entered at a command prompt at the correct
times.

The primary enhancement requests from the PSAS team are for a fully
automated simulation, for a faster simulation rate, for a more
complete rocket system model, and for a more complex physics model.

Automation of the simulation is highly desired. The current simulator
requires the operator to manually enter simulator commands during the
test flights. These commands can be initiated at the right time by
looking at messages from launch control and from the flight computer.

The current rate of simulation, at 20 frames per second, is much
slower than data rate produced by some of the rocket sensors. The
accelerometer produces 2,500 messages per second, and the gyroscope
produces 833 messages per second. Not every system needs to produce
data at the same rate, allowing systems to run at different rates
could be very useful.

Additional rocket sensors need to be implemented in the simulator.
Primarily GPS messages are missing from the current simulator.

The real rocket systems accept and respond to messages. A fully
functioning rocket simulator would respond to messages in the same way
that the real systems do. Ideally most of the real system code would
be used in the simulator, but this is likely to require extensive code
refactoring. Until that work is done, a simple model of message
responses will probably be good enough.

Improving the physics model to include full 3D motion simulation,
mach boundary conditions, wind, air density and perhaps even orbital
mechanics could be useful additions to support future work. I'm sure
that simple models, like that of wind, could be added with little work,
but the more complex models will probably have to wait.


New Simulator:

I will be writing the new simulator in C++ with an object oriented
design. The primary target platform is Debian GNU/Linux [5], but I
may also make it portable to MS Windows. I will use CppUnit [6] to
unit test the code. From my personal simulation projects I have a
code base for this type of work that I can pull from. I will be using
rapid-prototyping and rapid-development methodologies. I intend to
quickly build a basic working framework and then add more
functionality based on my remaining project time and the needs of the
PSAS software team.

There will be five sections of work; the configuration file, the
physics engine, the environment models, the rocket physics model,
and the rocket systems models.

The configuration file will contain all of the run-time settable
simulator parameters. This will include parameters like launch
location and elevation, sensor signal noise level, boost force and
much more. I will keep the file format simple initially, basically a
"name = value" format with sets of parameters marked with a header in
the form "[set name]".

The physics engine will be very basic and will tie together the
environment models and the rocket physics model. It will utilize the
forces reported by the environment models and the rocket physics
model to calculate the rocket's motion through 3D space. Due to the
tight schedule, I will probably keep the environment and rocket
physics models similar to their current implementation. These models
include gravity, ground, air-pressure, drag on the rocket body and on
the chutes.

Ideally I will spend most of my time working with the rocket systems
models. These models will simulate the rocket's sensors and actuators
and need to present a realistic and responsive interface to the
flight computer. To complete these models I will need to work closely
with the team to determine which systems must be implemented first,
and how they need to interact with the real rocket hardware.

The rocket will be simulated in real time. The physics engine and each
rocket system will run in its own thread with individual timing. The
timing rates will be specified in the configuration file and can be
adjusted to get the desired performance. I don't see any immediate
obstacles to simulating the rocket systems at full speed.


Future Work:

Continuing work on the simulator will be essential to meet the current
and future needs of the PSAS team. This work will focus on making the
simulator more generic, improving its models, and improving its use
for testing.

A more generic simulator would support advanced configuration from
files. As more elements of the environmental and rocket models can be
defined in configuration files, testing new designs under novel
conditions will become easier.

Improving the models of the environment and rocket will help to make
the simulated flights much more realistic and useful. Improvements
in the rocket model could include simulating what happens to the
rocket and its sensor data at supersonic speeds, at high altitudes,
or under unusual conditions such high spin. While simulating the
behavior of rocket systems, it would be ideal to use as much of the
actual rocket code as possible. Displaying the flight in 3D with the
actual amateur TV overlay could improve the diagnosis of problems
when things go wrong, and would be a great tool to use during public
demonstrations.

Turning the simulator into an fully automated rocket testing platform
will help to ensure the future success and safety of PSAS rocket
launches. It should be possible to run the simulation without having
access to any of the actual rocket hardware, as well as running it
with a selected subset. By using configuration files, it would be
great to be able to be able to script flight profiles, including
weather and flight glitches such as bad or missing sensor data.
Ultimately these flight profiles should be strung together for
automated rocket testing with logging, so that results can be
reported, and reproduced.


Bibliography:

1: http://psas.pdx.edu
2: http://psas.pdx.edu/RocketSim
3: http://cvs.psas.pdx.edu/c/tools/rocketsim/rocket.5c?rev=1.2
4: http://www.nickle.org
5: http://www.debian.org
6: http://cppunit.sourceforge.net/cgi-bin/moin.cgi