Wheeled Humanoid

Wheeled humanoid robots are being developed to perform complex manipulation and locomotion tasks, particularly in dynamic and unstructured environments. Current research focuses on improving control algorithms, often combining model-based approaches (like linear quadratic regulators) with data-driven methods (such as deep reinforcement learning and Gaussian processes) to handle uncertainties in object dynamics and environmental interactions. This work leverages high-fidelity simulations for real-to-sim adaptation, reducing the need for extensive real-world data collection. The resulting advancements have implications for applications such as warehouse automation, disaster response, and human-robot collaboration.