Register now After registration you will be able to apply for this opportunity online.
Advanced Digital Twin Replication of Forest Environments Using Real-World Data
This research addresses the limitations of conventional simulation environments in capturing the complexities of forest terrains for robotics and automation applications in forest management. Utilizing real-world LiDAR and RGBD data, the project aims to develop a high-fidelity meshing pipeline. These advanced simulation environments will significantly improve perception algorithms and navigation strategies, thereby enhancing resource allocation and emergency response capabilities in forest settings.
Keywords: 3D Meshing, Perception, Robotics and Automation
The rapid evolution of robotics and automation technologies has heralded unprecedented capabilities for forest management, from resource allocation to wildfire prevention. However, the success of these operations largely depends on the robot's ability to perceive and navigate in complex terrains. Conventional simulation environments fall short in capturing the intricate details present in natural settings, necessitating the creation of more realistic simulation scenarios based on real-world data.
This research aims to develop a pipeline for converting real-world LiDAR (Light Detection and Ranging) and RGBD (Red-Green-Blue-Depth) environmental data into high-fidelity simulation meshings. Such environments will be pivotal for enhancing perception algorithms and navigation strategies applicable to forest operations.
The rapid evolution of robotics and automation technologies has heralded unprecedented capabilities for forest management, from resource allocation to wildfire prevention. However, the success of these operations largely depends on the robot's ability to perceive and navigate in complex terrains. Conventional simulation environments fall short in capturing the intricate details present in natural settings, necessitating the creation of more realistic simulation scenarios based on real-world data.
This research aims to develop a pipeline for converting real-world LiDAR (Light Detection and Ranging) and RGBD (Red-Green-Blue-Depth) environmental data into high-fidelity simulation meshings. Such environments will be pivotal for enhancing perception algorithms and navigation strategies applicable to forest operations.
- WP1: Review existing simulation tools, meshing algorithms, and relevant data types to identify gaps and requirements.
- WP2: Meshing Pipeline; Develop a pipeline to convert LiDAR and RGBD data into simulation meshes, incorporating post-process optimization and global bundle adjustment for improved mesh registration.
- WP3: Simulation & Evaluation; Import meshes into Gazebo/Ignition and evaluates through benchmark tests focusing on perception and navigation.
- WP1: Review existing simulation tools, meshing algorithms, and relevant data types to identify gaps and requirements. - WP2: Meshing Pipeline; Develop a pipeline to convert LiDAR and RGBD data into simulation meshes, incorporating post-process optimization and global bundle adjustment for improved mesh registration. - WP3: Simulation & Evaluation; Import meshes into Gazebo/Ignition and evaluates through benchmark tests focusing on perception and navigation.
- Highly motivated for the topic.
- Basic knowledge of pointcloud registration, meshing.
- Good coding skills in both Python and C++.
- Highly motivated for the topic. - Basic knowledge of pointcloud registration, meshing. - Good coding skills in both Python and C++.
- Fan Yang: fanyang1@ethz.ch
- Pol Eyschen: peyschen@ethz.ch
- Fan Yang: fanyang1@ethz.ch - Pol Eyschen: peyschen@ethz.ch
Advanced Digital Twin Replication of Forest Environments Using Real-World Data.pdf
