DIRT

DIRT is a rover platform for autonomous soil sampling. The repository contains the rover control modules, navigation and sensing simulations, a Panda3D visualizer, CAD references, and machine-learning model helpers.

Features

• Autonomous waypoint navigation and mission sequencing
• Drill and spectrometer simulation
• GPS/navigation, distance-sensor, and vision-alignment modules
• Headless rover physics simulation
• Interactive Panda3D 3D rover viewer
• Raspberry Pi hardware integration points
• Optional local language-model loading helpers

Requirements

• Python 3.10 or newer
• NumPy
• OpenCV
• gpiozero
• scikit-learn
• smbus2
• pyserial
• Panda3D for the 3D viewer

Raspberry Pi GPIO support is installed only on Linux. ROS 2 support is optional.

Setup

Create and activate a virtual environment, then install the project dependencies:

py -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt

On Linux or Raspberry Pi, use the equivalent activation command:

source .venv/bin/activate
python -m pip install -r requirements.txt

Running the project

Main mission entry point

Run the complete simulated mission:

python main.py

Run a subsystem demonstration:

python main.py --test alignment
python main.py --test spectrometer
python main.py --test distance
python main.py --test navigation

The real-hardware path is controlled by config.py. Review the configuration and platform settings before connecting hardware.

Headless rover simulation

Run one simulation:

python simulation/dirt_sim_core.py

Select terrain and control profiles:

python simulation/dirt_sim_core.py \
  --terrain-mode rocky \
  --control-mode aggressive \
  --csv runs/rocky.csv

Run all terrain and control combinations:

python simulation/dirt_sim_core.py --batch --out sim_runs/

Generate deterministic random waypoints:

python simulation/dirt_sim_core.py --random-waypoints 6 --seed 3

Available terrain profiles are farmland, grassland, rolling, dunes, rocky, and testyard. Available control profiles are smooth, pursuit, and aggressive.

Panda3D 3D viewer

Launch the interactive viewer:

python simulation/rover_3d_panda.py

Useful options:

python simulation/rover_3d_panda.py --terrain-mode testyard
python simulation/rover_3d_panda.py --camera overhead
python simulation/rover_3d_panda.py --headless --iters 600
python simulation/rover_3d_panda.py --snapshot rover.png

Viewer controls include:

• Click: add a waypoint
• W / S: throttle
• A / D: steering
• M: manual mode
• 1: autonomous waypoint mode
• 2: mission mode
• C: change camera
• Space: pause
• R: reset
• Q: quit

Repository layout

Directory or file	Purpose
core/	Distance sensing and alignment-related core modules
navigation/	GPS waypoint and navigation logic
simulation/	Headless physics, alignment, spectrometer, and 3D simulations
spectrometer/	Spectrometer interfaces and simulation support
lm_models/	Optional language-model loading helpers
CAD/	Mechanical CAD files and assemblies
Sim_Data/	Simulation datasets and generated analysis data
tests/	Automated tests
config.py	Platform and hardware configuration
main.py	Main mission and subsystem-test entry point
requirements.txt	Python dependency declarations

Testing and validation

Run the test suite from the repository root:

python -m pytest

Compile-check all Python modules without running hardware or GUI code:

python -m compileall .

Some tests and demonstrations require optional hardware, camera, display, or Panda3D functionality. Use the headless simulation for environments without a display.

Data and generated files

Simulation runs write CSV logs containing pose, terrain, motor, power, and mission data. Prefer writing new runs into a separate output directory such as sim_runs/ rather than overwriting checked-in reference data.

Do not commit virtual environments, Python bytecode, API keys, or hardware credentials.