1.Task Description

Task:
In a simulation environment, implement the EngineAI T800 robot to imitate and execute actions from the "Professional Boxing Motion Dataset."

2.Official Support Materials

1.T800 URDF Model File
Provides the structural model of the T800 robot in URDF format, including joint definitions, constraints, and basic documentation.

2.Professional Boxing Motion Dataset
Motion capture dataset of human skeletal boxing movements (bvh / fbx format).
Boxing motion dataset remapped to the T800 skeleton (npy format).

3.Open-Source Base Code
Simulation environment deployment code (MuJoCo-based simulation environment).
Basic training code for imitation learning, including data loading, training, and simple example policies.

3.Participant-Provided Materials

1.Training and Inference Code
Implementation of action generation and control strategies, including core algorithms, training scripts, and inference deployment logic.

2.Computing Resources
Local or privately owned cloud computing resources required for training, to be provided by the participants themselves.

4.Submission Requirements

1.T800 Simulation Demo Video
A clear demonstration of the T800 robot performing boxing actions in the simulation environment, intended to showcase the robot’s movements and for evaluation purposes.

2.Reproducible Strategy Files and Configuration
Policy files and model parameter configurations that can be directly loaded into the official simulation deployment code to reproduce the submitted results.
Include necessary environment descriptions and operation instructions to allow evaluators to replicate and assess the performance in a standardized setting.