Robot Simulation and 3D Scene Intern

Job Responsibilities

• Assist in building robot simulation environments and digital twin scenarios to support robots in completing perception, navigation, and operational task simulation verification in complex indoor and outdoor environments.
• Participate in 3D modeling of roads, parks, underground garages, public areas, and other scenarios, scene editing, physical parameter configuration, and management of simulation assets.
• Assist in building a reusable simulation scenario asset library, and complete the maintenance of environmental and operational object models such as ground, walls, cabinets, doors, vehicles, obstacles, and cones.
• Responsible for configuring various sensors in the simulation environment and exporting simulation data, including modeling and debugging devices such as RGB cameras, depth cameras, LiDAR, IMU, and thermal imaging equipment.
• Complete the configuration of randomized parameters for simulation scenarios, adjusting variables such as lighting, materials, occlusion, object poses, friction, noise, and collisions to improve the algorithm's generalization ability in real-world scenarios.
• Assist in aligning the simulation robot with the real hardware parameters, standardizing model dimensions, joint limits, collision bodies, end tools, and sensor installation poses.
• Conduct simulation tests in cooperation with the perception, SLAM, motion control, and task planning teams, completing path verification, collision detection, extreme condition, and accessibility simulation experiments.
• Participate in 3D vision-related auxiliary work, including point cloud/depth map processing, 3D object annotation, camera calibration, pose estimation assistance, and 3D data quality inspection.
• Compare the differences between simulated data and real collected data, and assist in building dedicated datasets for recognition, scene understanding, and robot operation tasks.
• Organize and standardize the simulation toolchain, asset naming conventions, data export processes, and standardized test cases to improve the team's simulation and data production efficiency.

1. Basic General Ability (Base)

• Familiar with Linux/Ubuntu development environment, with basic engineering development and problem debugging skills;
• Proficient in Python, with basic C reading and writing development skills, able to independently write simulation scripts and data processing tools;
• Have experience with any of the following toolchains: robotic simulation, 3D modeling, or game engine.
• Master the concepts of basic 3D data formats: mesh, point cloud, depth map, pose, camera intrinsic and extrinsic parameters, coordinate transformation;
• Understand the robot coordinate system framework, be familiar with TF transformations, camera extrinsics, and 3D spatial geometric relationships;
• Possesses the ability to set up scenarios and can build operational and reusable simulation environments according to testing requirements;
• Habitually standardize engineering management, maintain models, scripts, configurations, and documentation, and ensure reproducibility in multi-person collaboration scenarios;
• Has a basic understanding of robotic systems and can collaborate in the development and testing of modules such as perception, SLAM, and motion control.

2. Specific technical directions (meeting any one is sufficient)

Track A: Robot Simulation and Digital Twins

• Familiar with any of the simulation platforms Isaac Sim, Gazebo, MuJoCo, PyBullet, Unity, Unreal, Webots, or CoppeliaSim; able to import and configure robots, sensors, and task object models; knowledgeable in at least one model file format such as URDF, MJCF, USD, FBX, or GLTF; able to configure simulation physical properties such as rigid bodies, colliders, joint constraints, inertia, friction, and sensors; capable of building testable simulation scenarios for robotic tasks, not limited to static 3D displays; experience in simulation projects involving robotic arms, mobile robots, or humanoid robots is preferred.

Track B: Complex Scene Modeling and Asset Management

• Possesses hands-on experience in building 3D scenes and can complete modeling and organization of assets for indoor spaces, garages, campuses, cabinets, obstacles, and other scenes; Proficient in using any 3D modeling tools such as Blender, SketchUp, Maya; Able to recreate realistic simulation scenes based on real photos, CAD, and point cloud data; Capable of building a standardized reusable asset library, unifying model naming, hierarchy, materials, dimensions, and colliders; Understands the impact of lighting, reflections, occlusions, and materials on visual sensor simulation effects; Experience in programmatic random scene generation and synthetic data creation is a plus.

Track C: 3D Vision and Sensor Data Processing

• Proficient in using visual data processing tools such as OpenCV, Open3D, PCL, NumPy, and PyTorch3D; familiar with camera imaging models, depth maps, point clouds, calibration, 3D projection, and coordinate transformation principles; able to handle RGB-D, LiDAR, and multi-camera simulation data; assist in object pose estimation, scene geometry analysis, and robot operable area determination; collaborate with algorithm engineers to generate, annotate, quality-check simulation data, and analyze differences between virtual and real data; experience in robot vision, 3D reconstruction, or NeRF/Gaussian Splatting related study or projects is preferred.

Preferred Qualifications

• Participated in work related to building a robot simulation platform from scratch;
• Has practical project experience with Isaac Sim, Omniverse, Gazebo, MuJoCo, Unity, and Unreal;
• Have experience in simulation development of robotic arms, quadrupeds, humanoids, mobile robots, or autonomous driving.
• In the simulation, practical tasks such as grasping, pushing doors, pressing buttons, transporting, and tool operation have been completed and verified.
• Independently complete modeling of large and complex scenes such as garages, parks, factories, and warehouses;
• Proficient in domain randomization, synthetic dataset generation, and Sim2Real transfer-related technologies;
• Familiar with ROS / ROS2, capable of integrating the simulation environment into the robot algorithm development pipeline;
• Master the complete process of simulation configuration and calibration for cameras, LiDAR, and IMU.
• Exposure to 3DGS, NeRF, point cloud registration, SLAM map reconstruction, and 3D reconstruction related content;
• Have experience in robotics competitions, open-source simulation projects, and complete projects related to digital twins;
• Able to use AI tools such as Claude Code and Codex to assist in script and simulation development.