Contact Modeling

Contact modeling in robotics focuses on accurately representing and predicting the interactions between robots and objects, crucial for tasks like manipulation and grasping. Current research emphasizes developing more efficient and robust models, moving away from complex complementarity formulations towards differentiable and optimization-based approaches, often incorporating generative models or neural networks like transformers to handle diverse object shapes and contact scenarios. This improved modeling enables more accurate and efficient planning and control algorithms for robots interacting with the physical world, impacting fields such as dexterous manipulation, nonprehensile manipulation, and grasp generation.