. Apply VLM/VLA for robot locomotion, manipulation, and complex task planning. . Utilize Decision Tree deployment methodologies to convert high-level instructions into specific robot action commands. . Develop RL-based control strategies for dual-arm coordinated manipulation, including synchronization, trajectory planning, and collision avoidance. . Select and optimize RL algorithms within the digital twin environment to improve training efficiency. . Integrate force or tactile sensors feedback into the RL training loop to enable force or torque-based control in digital twin platform. . Standardize the Input/Output structure of RL models, including pose/sensor data representation and command formats. . Perform Real-to-Sim calibration to align dynamics and sensing parameters, and implement Sim-to-Real transfer strategies to reduce the reality gap . Design and build dual-arm Digital Twins environments for VLM/RL model training, testing, and validation. . Create API interfaces to integrate the virtual training environment with ROS 2 for VLM/RL control implementation. . Collaborate with other teams to ensure accurate modeling of kinematics, dynamics, sensors, and actuators in the Digital Twin. . Stay updated and apply the latest tools and methodologies in Meta-Learning, Domain Randomization, and Digital Twin technologies to improve the model's generalization abilities.
