AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning
This work addresses a critical challenge in robotics for real-world applications, offering an incremental improvement by enhancing environment diversity and learning methods for adaptive manipulation.
The paper tackles the problem of robot manipulation of articulated objects with hidden internal states, such as locks or joint constraints, by introducing a new simulation environment with diverse object categories and an adaptive policy learning pipeline. The method is validated through simulation and real-world experiments, showing effectiveness in handling complex manipulation tasks.
Articulated object manipulation is a critical capability for robots to perform various tasks in real-world scenarios. Composed of multiple parts connected by joints, articulated objects are endowed with diverse functional mechanisms through complex relative motions. For example, a safe consists of a door, a handle, and a lock, where the door can only be opened when the latch is unlocked. The internal structure, such as the state of a lock or joint angle constraints, cannot be directly observed from visual observation. Consequently, successful manipulation of these objects requires adaptive adjustment based on trial and error rather than a one-time visual inference. However, previous datasets and simulation environments for articulated objects have primarily focused on simple manipulation mechanisms where the complete manipulation process can be inferred from the object's appearance. To enhance the diversity and complexity of adaptive manipulation mechanisms, we build a novel articulated object manipulation environment and equip it with 9 categories of objects. Based on the environment and objects, we further propose an adaptive demonstration collection and 3D visual diffusion-based imitation learning pipeline that learns the adaptive manipulation policy. The effectiveness of our designs and proposed method is validated through both simulation and real-world experiments. Our project page is available at: https://adamanip.github.io