A railway maintenance device configured for autonomous or semi-autonomous operation in a rail environment is provided. The device can process image data to move about the rail environment and perform one or more actions in the rail environment. The actions can include one or more actions related to decoupling and/or attaching rail vehicles, and can be implemented by performing three-dimensional image processing. The device can be configured to move with any movement of a rail vehicle on which one or more actions are being performed. In an alternative embodiment, the various components configured to perform the action are implemented at a stationary location with respect to a rail line.