New AI Makes Robots Better at Handling Ropes and Cables

Researchers present BM2PC, a Bayesian Multimodal MPC framework for agile manipulation of deformable linear objects (DLOs). It formulates multimodal planning as a set optimization problem balancing control quality and behavioral diversity, with an efficient incremental update and an execution dynamics model for real-world fast manipulation.

In simulation and hardware tests on rope tracking and obstacle avoidance, plus simulated disturbance rejection, BM2PC generalized better to unseen trajectories than a learning-based baseline without task-specific policy training or fine-tuning. Real-world experiments showed accurate long-horizon tracking, reducing RMSE by 44.0% on the longest task. In obstacle avoidance, BM2PC achieved comparable performance to existing multimodal MPC methods while reducing planning time by 29% to 66%.