This article presents the results of the experimental campaign of the EU-H2020 funded project DexROV. Its goal is to enable the remote operation of an underwater vehicle-manipulator system (UVMS) via satellite communication to execute intervention tasks in an Oil and Gas industry scenario. More in detail, this work focuses on the overall control architecture that has been deployed and tested during the DexROV June 2018 experimental campaign held in Marseilles, France. The motion controller relies on a task-priority inverse kinematics algorithm that allows handling several control objectives and performing them simultaneously exploiting the system redundancy. Classical inverse kinematics algorithms have been properly extended to handle also set-based tasks, or inequality constraints, useful for the definition of several safety tasks such as joint limits or virtual walls. The control architecture also includes an admittance loop that exploits the force/torque measurements of a wrench sensor in order to make the system compliant to undesired external forces and unexpected collisions with the environment. Preparatory simulations and experimental trials are presented for proving the effectiveness of the deployed architecture in a ``turn the valve'' operation.

In this paper we present an architecture for the operation of an assistive robot finally aimed at allowing users with severe motion disabilities to perform manipulation tasks that may help in daily-life operations. The robotic system, based on a lightweight robot manipulator, receives high level commands from the user through a Brain-Computer Interface based on P300 paradigm. The motion of the manipulator is controlled relying on a closed loop inverse kinematic algorithm that simultaneously manages multiple set-based and equality-based tasks. The software architecture is developed relying on widely used frameworks to operate BCIs and robots (namely, BCI2000 for the operation of the BCI and ROS for the control of the manipulator) integrating control, perception and communication modules developed for the application at hand. Preliminary experiments have been conducted to show the potentialities of the developed architecture.

In this paper, we present DexROV, an EC Horizon 2020-funded project that proposes to implement novel operation strategies for underwater semiautonomous interventions. These costly and demanding operations are increasingly performed by remotely operated vehicles (ROVs), contributing to risk cutting for human divers. However, ROV operations require offshore structures, hosted on a support vessel with a crew of a significant amount of personnel necessary to properly handle and operate the robotic platform. One of the key goals of DexROV is to delocalize on shore the manned support as much as possible, reducing the crew onboard the support vessel and consequently the costs and risks of the whole operation. The control center is located onshore, far from the actual operation location. Operators interact with the ROV through a simulation environment that exploit 3D models of the environment built online relying on the perception and modeling capabilities of the robotic system and transmitted via satellite communication. Currently, ROVs lack the dexterous capabilities needed to perform many kinds of operations, for which human divers are still necessary. DexROV addresses this problem, equipping the ROV with two 6 DOF (degrees of freedom) dexterous manipulators with anthropomorphic end-effectors and providing semiautonomous capabilities. The control will rely on a multitask priority approach that will help the operator to focus on the main operation, leaving the low-level tasks to be autonomously performed by the ROV.

In this paper we present a control architecture for a robotic manipulator finally aimed at helping people with severe motion disabilities in performing daily-life operations, such as manipulating objects or drinking. The proposed solution allows the user to focus the attention only on the operational tasks, while all the safety-related issues are automatically handled by the developed control architecture. The user commands the manipulator sending high-level commands via a P300-based Brain-Computer Interface. A perception module, relying on a RGB-D sensor, continuously detects and localizes the objects in the scene, tracking the position of the user and monitoring the environment for identifying static and dynamic obstacles, e.g., a person entering in the scene. A lightweight manipulator is controlled relying on a Task Priority Inverse Kinematics algorithm that handles task hierarchies composed of equality-based and set-based tasks, including obstacle avoidance and joint mechanical limits. The paper describes the overall architecture and the integration of the implemented software modules, that are based on common frameworks and software libraries such as ROS, BCI2000, OpenCV and PCL. Experimental results on a use case scenario using a Kinova 7DOFs Jaco robot helping a user to perform drinking and manipulation tasks show the effectiveness of the developed control architecture.

Task-priority inverse kinematics is a popular motion control algorithm which efficiently handles redundancy in robot manipulators. It has been recently extended in order to handle also set-based control objectives or inequality constraints. As any local motion planner it is prone to the occurrence of local minima. This work further extends set-based inverse kinematics by adding a motion planner in order to avoid such occurrence. Motion planners are usually computationally heavy especially in their eventual implementation with a task-priority architecture. To reduce this issue, the planner is implemented as a sampling-based algorithm which works in the reduced-dimensionality of the robot workspace applying Cartesian constraints only. The output trajectory is then checked against the inverse kinematics algorithm exploiting the redundancy and verifying the fulfillment of the joint-based task constraints. During the motion, inverse kinematics is then used also in real-time to ensure a reactive behavior to address, e.g., mismatch between the a-priori information and real-time perception acquisition. Also, the motion planner runs in background to adapt to changes in the environment or to accommodate incremental mapping. Comparison with alternative approaches are investigated and discussed. The most promising method is validated first in hundreds of numerical simulations to provide a solid statistical analysis and then experimentally with a Kinova Jaco2 7 DOFs manipulator equipped with an RGB-D sensor. Note to Practitioners—In this work we propose a motion planning algorithm that allows to effectively deploy robot manipulators in partially unstructured environments. It is structured in two layers: a Cartesian space global planner guarantees the feasibility of the trajectory with respect to potential obstacles in the environment, while a reactive local planner checks the feasibility at joint level. This architecture allows to define all the safety constraints both at Cartesian and joint space needed to operate in unstructured environments, while keeping the computation time lower with respect to standard approaches that define all the constraints in an offline global planner. The reduced computation time allows to effectively perform real-time replanning operations when needed, making the robotic system capable of adapting to a dynamic environment and suitable for sharing its workspace with human operators.