IROS26

Planned Release for IROS 2026 PlantShade: Predicting Plant Shadows for Lighting-Aware Robotic Agricultural Operation.

UE5.6 source project (some assets excluded due to licenses). The packaged simulator and instructions are planned.

Before the planned packaged simulator is released, see the following files for implementations

• Procedural model loading entry point: ShadowPlant_BP.uasset
• Procedural model loading cpp entry point: ROS2LoadModel.h and UDynamicPlantMesh.h
• Shadow post-process shader: Shadow_MatPP.uasset
• Shadow camera render: Robot_BP.uasset

Datasets:

• Main Dataset (~70 GB): https://cornell.box.com/s/kkm168x304hwjc5m8h7722mrvaqwk1c3
• Full combination (~70 GB): https://cornell.box.com/s/1t2ped3uyy0b6pj0xi9a68pfq5noz4uw

Additional Resources for data collection:

• Packaged Simulator: TODO
• Helios Fork with mesh optimization https://github.com/XingjianL/Helios/tree/master/projects/procedural_model
• ROS data collection process https://github.com/XingjianL/UE5Sim_colcon_ws/tree/tomato/src/shadows

Instructions with Editor
Note: The installation and data generation process will be significantly more convoluted than described below due to multiple project requirements across possibly multiple environments.

Environment: Windows 11 with WSL

• Windows 11 for UE project
• WSL for ROS and Helios

1. Source the ROS environment
2. Start the rosbridge server ros2 launch tomato_xarm6 upside_down_xarm6_moveit.launch.py
3. Start the UE5 simulator and change the environment to ShadowDataCollect
4. Run the Helios project procedual_model after build, a plant model should appear in the UE5 environment.
5. Run the data collection script shadow_data_collect.sh