This paper presents the results of the experimental tests performed to validate the functionality of a variable pitch system (VPS), designed for pitch attitude control of the novel underwater robotic vehicle explorer UX-1. The VPS is composed of a mass suspended from a central rod mounted across the hull. This mass is rotated around the transverse axis of the vehicle in order to perform a change in the inclination angle for navigation in vertical mine shafts. In this work, the equations of motion are first derived with a quaternion attitude representation, and are then extended to include the dynamics of the VPS. The performance of the VPS is demonstrated in real underwater experimental tests that validate the pitch angle control independently, and coupled with the heave motion control system.

This paper proposes a manipulability optimization control of a 7-DoF robot manipulator for Robot-Assisted Minimally Invasive Surgery (RAMIS), which at the same time guarantees a Remote Center of Motion (RCM). The first degree of redundancy of the manipulator is used to achieve an RCM constraint, the second one is adopted for manipulability optimization. A hierarchical operational space formulation is introduced to integrate all the control components, including a Cartesian compliance control involving the main surgical task, a first null-space controller for the RCM constraint, and a second null-space controller for manipulability optimization. Experiments with virtual surgical tasks, in an augmented reality environment, were performed to validate the proposed control strategy using the KUKA LWR 4 +. The results demonstrate that end-effector accuracy and RCM constraint can be guaranteed, along with improving the manipulability of the surgical tip.

This paper presents a newly conceived planning algorithm that is based on the introduction of motion primitives in RRT*. Online computational complexity of RRT* is greatly reduced by pre-computing the optimal constrained trajectories joining pairs of starting and destination configurations in a grid space, while taking into account vehicle motion constraints in the planning task. A numerical example shows the effectiveness of the algorithm.

In this work we present a method to detect and estimate the three-dimensional pose of planar and textureless objects placed randomly on a conveyor belt or inside a bin. The method is based on analysis of single 2D images acquired by a standard camera. The algorithm exploits a template matching method to recognize the objects. A set of pose hypotheses are then refined and, based on a gradient orientation scoring, the best object to be manipulated is selected. The method is flexible and can be used with different objects without changing parameters since it exploits a CAD model as input for template generation. We validated the method using synthetic images. An experimental setup has been also designed using a fixed standard camera to localize planar metal objects in various scenarios.

Although force control algorithms have been studied for three decades, this technology is not largely exploited in industry yet. The present paper proposes a position-based adaptive force control strategy, that relies on a novel method for the on line estimation of the environment stiffness. The control design is targeted to industrial controller structures and it is theoretically proven to be robust to time varying estimation errors of the environment stiffness and joint friction disturbances. The estimation algorithm succeeds in identifying the environment stiffness even in presence of geometrical irregularities of the contact surface during motion. The identification and control approaches are experimentally validated on an industrial robot equipped with a force sensor.

Classical control approaches to robot force control have been extensively addressed by research in the last decades and are now considered a paradigm when dealing with force control for industrial robots. With this respect, the present paper exploits the capability of state-of-the-art Quadratic Programming (QP) solvers to specify a simple and intuitive constraint-based optimization strategy aiming at improving closed-loop performance of a classical force controller, such as the implicit force control with pure integral action for a position-controlled manipulator in contact with a compliant environment. The effectiveness of the proposed control strategy is experimentally validated on an industrial robot equipped with a force sensor.

To perform many common industrial robotic tasks, e.g. deburring a work-piece, in small and medium size companies where a model of the work-piece may not be available, building a geometrical model of how to perform the task from a data set of human demonstrations is highly demanded. In many cases, however, the human demonstrations may be sub-optimal and noisy solutions to the problem of performing a task. For example, an expert may not completely remove the burrs that result in deburring residuals on the work-piece. Hence, we present an iterative algorithm to estimate a noise-free geometrical model of a work-piece from a given dataset of profiles with deburring residuals. In a case study, we compare the profiles obtained with the proposed method, nonlinear principal component analysis and Gaussian mixture model/Gaussian mixture regression. The comparison illustrates the effectiveness of the proposed method, in terms of accuracy, to compute a noise-free profile model of a task.

Dynamic Movement Primitives (DMPs) are a common method for learning a control policy for a task from demonstration. This control policy consists of differential equations that can create a smooth trajectory to a new goal point. However, DMPs only have a limited ability to generalize the demonstration to new environments and solve problems such as obstacle avoidance. Moreover, standard DMP learning does not cope with the noise inherent to human demonstrations. Here, we propose an approach for robot learning from demonstration that can generalize noisy task demonstrations to a new goal point and to an environment with obstacles. This strategy for robot learning from demonstration results in a control policy that incorporates different types of learning from demonstration, which correspond to different types of observational learning as outlined in developmental psychology.

This paper investigates the use of second-order methods to solve Markov Decision Processes (MDPs). Despite the popularity of second-order methods in optimization literature, so far little attention has been paid to the extension of such techniques to face sequential decision problems. Here we provide a model-free Reinforcement Learning method that estimates the Newton direction by sampling directly in the parameter space. In order to compute the Newton direction we provide the formulation of the Hessian of the expected return, a technique for variance reduction in the sample-based estimation and a finite sample analysis in the case of Normal distribution. Beside discussing the theoretical properties, we empirically evaluate the method on an instructional linear-quadratic regulator and on a complex dynamical quadrotor system.

This paper investigates the use of policy gradient techniques to approximate the Pareto frontier in Multi-Objective Markov Decision Processes (MOMDPs). Despite the popularity of policy-gradient algorithms and the fact that gradient-ascent algorithms have been already proposed to numerically solve multi-objective optimization problems, especially in combination with multi-objective evolutionary algorithms, so far little attention has been paid to the use of gradient information to face multi-objective sequential decision problems. Three different Multi-Objective Reinforcement-Learning (MORL) approaches are here presented. The first two, called radial and Pareto following, start from an initial policy and perform gradient-based policy-search procedures aimed at finding a set of non-dominated policies. Differently, the third approach performs a single gradient-ascent run that, at each step, generates an improved continuous approximation of the Pareto frontier. The parameters of a function that defines a manifold in the policy parameter space are updated following the gradient of some performance criterion so that the sequence of candidate solutions gets as close as possible to the Pareto front. Besides reviewing the three different approaches and discussing their main properties, we empirically compare them with other MORL algorithms on two interesting MOMDPs.

In an industiral scenario the capability to detect and track human workers entering a robotic cell represents a fundamental requirement to enable safe and efficient human-robot cooperation. This paper proposes a new approach to the problem of Human Detection and Tracking based on low-cost commercial RGB surveillance cameras, image warping techniques, computer vision algorithms, efficient data structures such as k-dimensional trees and particle filtering. Results of several validation experiments are presented.

In the past few years the need for more flexibility in industrial production has implied a growing attention towards scenarios where humans work directly in touch with robots. In order to allow safe human-robot interaction, a methodology to evaluate the severity of an impact between a human worker and an industrial robot, based on related work on injury knowledge in human-robot contacts and relying on information coming from different exteroceptive sensors, has been developed in this paper. On the basis of this severity evaluation, the robot controller enforces a suitable safety-oriented strategy, ranging from on-path speed reduction to task-consistent evasive motion and protective stop. The safety evaluation methodology has been implemented in a dedicated software component, integrated with a video surveillance system and with the real time robot controller to obtain a complete HW/SW architecture named “Safety Controller”. The system has been validated on an ABB IRB140 robot.

This paper investigates the use of policy gradient techniques to approximate the Pareto frontier in Multi-Objective Markov Decision Processes (MOMDPs). Despite the popularity of policy gradient algorithms and the fact that gradient ascent algorithms have been already proposed to numerically solve multi-objective optimization problems, especially in combination with multi-objective evolutionary algorithms, so far little attention has been paid to the use of gradient information to face multi-objective sequential decision problems. Two different Multi-Objective Reinforcement-Learning (MORL) approaches, called radial and Pareto following, that, starting from an initial policy, perform gradient-based policy-search procedures aimed at finding a set of non-dominated policies are here presented. Both algorithms are empirically evaluated and compared to state-of-the-art MORL algorithms on three MORL benchmark problems.

The main purpose of this paper is to present an implicit force control scheme for 6 DoF industrial robots, whose compliance model takes into account both joint and link elasticities. The system composed by the robot and the environment is modeled by means of a simple equivalent elasticity at the end effector. A method to avoid limit cycles due to friction during force control applications is then proposed. The compliance model and the force control are experimentally validated on a industrial robot equipped with a force sensor.

Industrial robots are today separated from human workers by means of safety barriers, that protect humans from the risk of collisions. This separation has a clear negative influence on diffusion of robotic technology in shopfloors. On the other hand the removal of protective barriers gives rise to safety issues, that can be addressed with a combination of approaches, including sensor based reactive control. In this paper a distributed proximity sensor, to be mounted on the links of the manipulator, is presented. The optimal placement of the spots of such sensor is discussed, taking into account detection capabilities and safety enhancement. Experiments developed on an ABB IRB 140 robot using off-the-shelf infrared distance sensors as spots are presented.

This work investigates the way humans plan their paths in a goal-directed motion. The person can be viewed as an optimal controller that plans the path minimizing a certain (unknown) cost function. Taking this viewpoint, the problem can be formulated as an inverse optimal control one, i.e., starting from control and state trajectories we want to figure out the cost function used by a person while planning the path. To test the envisaged ideas, a set of walking paths of different volunteers were recorded using a motion capture facility. The collected data have been used to compare a solution to the inverse control problem coming from the literature to a novel one. The obtained results, ranked using the discrete Fréchet distance, show the effectiveness of the proposed approach.

In the last decade, policy gradient methods have significantly grown in popularity in the reinforcement--learning field. In particular, they have been largely employed in motor control and robotic applications, thanks to their ability to cope with continuous state and action domains and partial observable problems. Policy gradient researches have been mainly focused on the identification of effective gradient directions and the proposal of efficient estimation algorithms. Nonetheless, the performance of policy gradient methods is determined not only by the gradient direction, since convergence properties are strongly influenced by the choice of the step size: small values imply slow convergence rate, while large values may lead to oscillations or even divergence of the policy parameters. Step--size value is usually chosen by hand tuning and still little attention has been paid to its automatic selection. In this paper, we propose to determine the learning rate by maximizing a lower bound to the expected performance gain. Focusing on Gaussian policies, we derive a lower bound that is second--order polynomial of the step size, and we show how a simplified version of such lower bound can be maximized when the gradient is estimated from trajectory samples. The properties of the proposed approach are empirically evaluated in a linear--quadratic regulator problem.

Research in the field of human kinematic analysis has gained interest in recent years and has fostered new ideas and expectations. Next generation manipulators are expected to resemble a human-like behaviour at kinematic level, in order to avoid any unease or discomfort (like fear or shock) to the nearby humans. In this work, a kinematic experimental approach to study and synthesize the motion of the human arm is presented. In particular, the proposed scenario will be used to study how humans exploit the kinematic redundancy of their arms, for a future use in a robotic controller. A simple, yet accurate, method for human-like redundancy resolution in robotic manipulators is developed and verified.

Removing the safety fences that separate humans and robots, to allow for an effective human-robot interaction, requires innovative safety control systems. An advanced functionality of a safety controller might be to detect the presence of humans entering the robotic cell and to estimate their intention, in order to enforce an effective safety reaction. This paper proposes advanced algorithms for cognitive vision, empowered by a dynamic model of human walking, for detection and tracking of humans. Intention estimation is then addressed as the problem of predicting online the trajectory of the human, given a set of trajectories of walking people learnt offline using an unsupervised classification algorithm. Results of the application of the presented approach to a large number of experiments on volunteers are also reported.

Robust controllers for robot manipulators ensure stability of the closed-loop system, even if only partial knowledge of the dynamic model of the manipulator is available. Existing derivations of robust-control laws, while guaranteeing the stability result, present an undesired dependence of the robust-control term on the gains of the controller for the nominal system. This dependence forces larger robust-control terms when the nominal control gains are large. Based on a structured representation of the model uncertainty, this paper proposes a derivation of the robust-control law, where these limitations are removed. Experimental results on the COMAU SMART 3S industrial robot in a 3-degree-of-freedom (DOF) configuration confirm the advantages of the proposed controller.

Robust controllers for robot manipulators ensure stability properties of the closed loop system, even if only partial knowledge of the dynamic model of the manipulator is available. Existing derivations of robust control laws, while guaranteeing the stability result, present an undesired interaction between the gains of the controller of the nominal system and the robust control term. Based on a structured representation of the model uncertainty, this paper presents a derivation of the robust control law where these limitations are removed. A case study is discussed to show the benefits of the proposed approach. New insight in the robust control problem for more general mechanical systems might arise from structuring the model uncertainty as proposed in this paper.

Visual servoing of eye-in-hand flexible manipulators is addressed in this paper. Dynamic effects of both the rigid and the flexible motion of the manipulator are fully taken into account in a control solution where the two-time scale nature of the problem is exploited. The visual information is used in the "slow" subsystem for a task-space-oriented control law, where computationally expensive operations, such as inverse and time derivative of the Jacobian, are avoided. A constructive proof of stability of this control scheme, based on Lyapunov theory, is also presented. The effectiveness of the proposed controller is shown by means of a numerical simulation concerning a trajectory tracking problem. Some experimental results finally demonstrate the precision enhancement achieved by the proposed algorithm on a single-link flexible manipulator