A Modular ROS-based dVRK Teleoperation Controller Architecture

Abstract

The da Vinci Research Kit (dVRK) [1], a hardware and software stack for research development, implements the teleoperation control of the Patient Side Manipulators (PSMs) using the Master Tool Manipulators (MTMs), with the original capabilities of the da Vinci Surgical System such as input device clutching, rotated vision perspective control, and input output position scaling. A recent research trend in automation and learning [2], uses 3D simulators such as AMBF [3], SurRol [4], and UnityFlexML [5]. These simulators re-implement the teleoperation controller and kinematics solver due to the non-modularity of the dVRK software (Fig. 1a). We ar- gue that future dVRK research requires an architecture, that supports testing algorithms in simulation and real hardware, with a singular teleoperation controller and kinematics. This in turn will provide a robust robot test workflow and standardized algorithm comparisons and benchmarks. Another research trend is the use of alternative input devices to control simulated PSMs. AMBF incorporates a large variety of input device options. Additionally, there has been work in using the MTM to teleoperate in- dustrial robots [6]. In all of these cases, the teleoperation controller and kinematics are re-implemented. To the best of our knowledge, there does not exist a formal definition of the dVRK teleoperation interfaces. We propose a modular dVRK teleoperation controller architecture, that is based on the Robot Operating Sys- tem (ROS), with the original capabilities and further improvements such as: • A lightweight and minimal dependency software implementation with a modular kinematics solver • Defined input interface to integrate a variety of input devices • Defined output interfaces that integrates to existing dVRK simulations and hardware