DEIMOS Autonomy Stack

Drop-in autonomy layer for Space Concordia's DEIMOS rover, targeting URC 2026 Finals.

Implements the architecture described in docs/AUTONOMY_PROPOSAL.md:

• SLAM: FAST-LIO2 (Orin NX path) or slam_toolbox 2D + ZED SDK VIO (Orin Nano path)
• Sensor fusion: robot_localization dual-EKF (local + global), fusing 4 onboard IMUs + GPS + SLAM odom
• Global planning: USGS DEM A* over MDRS terrain (WVU-style pre-loaded elevation map)
• Local planning: Nav2 Regulated Pure Pursuit
• Final approach: ArUco visual servo action server
• Object detection: YOLOv8n TensorRT (FP16 on DLA / INT8 on GPU)
• Mission orchestration: Goal sequencer state machine

Modeled on the publicly documented stacks of WVU Team Mountaineers (2025 URC top autonomy scorer) and AGH Space Systems (2024 URC champion).

---

Quick Start

Prerequisites

• Ubuntu 22.04
• ROS 2 Humble
• TensorRT 8.5+ (Jetson)
• Python 3.10+

Install dependencies

sudo apt update
sudo apt install \
    ros-humble-robot-localization \
    ros-humble-slam-toolbox \
    ros-humble-pointcloud-to-laserscan \
    ros-humble-octomap-server \
    ros-humble-vision-msgs \
    ros-humble-nav2-bringup \
    ros-humble-nav2-msgs \
    ros-humble-topic-tools

# Python deps for DEM planner
pip install rasterio pyproj numpy opencv-python

# Optional: ZED SDK (Stereolabs official installer)
# Optional: TensorRT + pycuda for YOLO
pip install pycuda

Build into existing workspace

Drop src/deimos_autonomy next to the existing src/ packages from robot-repo-ros2.

cd ~/your_ros2_ws/src
ln -s /path/to/deimos-autonomy/src/deimos_autonomy .
cd ..
colcon build --symlink-install --packages-select deimos_autonomy
source install/setup.bash

Run

# Orin Nano profile (slam_toolbox 2D + ZED VIO)
ros2 launch deimos_autonomy autonomy.launch.py compute_profile:=nano_8gb

# Orin NX profile (FAST-LIO2 LiDAR-inertial)
ros2 launch deimos_autonomy autonomy.launch.py compute_profile:=nx_16gb

# With Gazebo simulation
ros2 launch deimos_autonomy autonomy.launch.py compute_profile:=nano_8gb use_sim_time:=true

Pre-flight diagnostic

Before integrating, run the read-only diagnostic on your rover:

python3 scripts/deimos_preflight.py

It validates ROS2 topics, TF tree, CAN bus, Ouster reachability, ZED2 USB, power mode, RAM/disk headroom, and JetPack version. Pass/warn/fail report with actionable fixes. Sample report at docs/SAMPLE_PREFLIGHT_REPORT.txt.

---

Topic graph

Consumed (existing topics, see robot-repo-ros2)

Topic	Type	Source
/ouster/points	sensor_msgs/PointCloud2	ouster_ros (lidar_launch.py)
/ouster/imu	sensor_msgs/Imu	ouster_ros
/zed/zed_node/rgb/image_rect_color	sensor_msgs/Image	zed_wrapper
/zed/zed_node/imu/data	sensor_msgs/Imu	zed_wrapper
/zed/zed_node/odom	nav_msgs/Odometry	ZED SDK Positional Tracking (Nano profile)
/fix	sensor_msgs/NavSatFix	ublox_gps_node
/aruco_detections	aruco_opencv_msgs/ArucoDetection	ros_aruco_opencv

Published (this package)

Topic	Type	Producer
/odometry/dlio	nav_msgs/Odometry	FAST-LIO2 or ZED-VIO bridge
/odometry/filtered	nav_msgs/Odometry	EKF local
/odometry/filtered_map	nav_msgs/Odometry	EKF global
/odometry/gps	nav_msgs/Odometry	navsat_transform_node
/map	nav_msgs/OccupancyGrid	slam_toolbox (Nano only)
/scan	sensor_msgs/LaserScan	pointcloud_to_laserscan (Nano only)
/plan	nav_msgs/Path	dem_global_planner
/traversability	nav_msgs/OccupancyGrid	dem_global_planner (debug)
/yolo/detections	vision_msgs/Detection2DArray	yolo_trt_node
/cmd_vel	geometry_msgs/Twist	Nav2 controller / aruco_action_server
/mission/state	std_msgs/String	goal_sequencer

---

File map

deimos_autonomy/
├── docs/
│   ├── AUTONOMY_PROPOSAL.md          ← Architecture overview, design rationale, references
│   └── COMPETITION_RUNBOOK.md        ← Deployment, troubleshooting, fallback procedures
├── src/deimos_autonomy/
│   ├── package.xml
│   ├── setup.py
│   ├── launch/
│   │   ├── autonomy.launch.py        ← Master launch with compute_profile switching
│   │   ├── fast_lio.launch.py        ← FAST-LIO2 (NX path)
│   │   ├── slam_toolbox_2d.launch.py ← slam_toolbox + ZED VIO (Nano path)
│   │   ├── ekf.launch.py             ← Dual-EKF + navsat_transform
│   │   └── nav2.launch.py            ← Nav2 lifecycle nodes
│   ├── config/
│   │   ├── fast_lio_ouster.yaml      ← Ouster OS1 tuning
│   │   ├── slam_toolbox_2d.yaml
│   │   ├── ekf_local.yaml            ← 4-IMU fusion
│   │   ├── ekf_global.yaml           ← + GPS
│   │   ├── navsat_transform.yaml
│   │   ├── nav2_params.yaml          ← Nav2 + costmaps
│   │   └── nav2_bt_navigate.xml      ← Behavior tree
│   └── deimos_autonomy/
│       ├── aruco_action_server.py    ← Final-approach visual servo
│       ├── yolo_trt_node.py          ← TRT YOLOv8n inference
│       ├── dem_global_planner.py     ← USGS DEM A*
│       └── goal_sequencer.py         ← Mission state machine
└── scripts/
    ├── train_yolo_urc.py             ← Fine-tune YOLOv8n on URC objects (DGX Spark)
    ├── build_yolo_trt.sh             ← ONNX -> TRT engine (Jetson)
    ├── download_mdrs_dem.sh          ← USGS National Map fetch
    └── record_demo_bag.sh            ← Record sample data for Gazebo replay

---

License

PolyForm Noncommercial 1.0.0. See LICENSE. Free for research, education, and noncommercial use. Commercial use requires a separate license — contact pyloxsystems@gmail.com.

---

Author

Emilio Girard / Pylox Forge — pyloxsystems@gmail.com