This study proposes a method based on reinforcement learning (RL) for point-to-point and dynamic reference position tracking control of a planar cable-driven parallel robots, which is a multi-input multi-output system (MIMO). The method eliminates the use of a tension distribution algorithm in controlling the system’s dynamics and inherently optimizes the cable tensions based on the reward function during the learning process. The deep deterministic policy gradient algorithm is utilized for training the RL agents in point-to-point and dynamic reference tracking tasks. The performances of the two agents are tested on their specifically trained tasks. Moreover, we also implement the agent trained for point-to-point tasks on the dynamic reference tracking and vice versa. The performances of the RL agents are compared with a classical PD controller. The results show that RL can perform quite well without the requirement of designing different controllers for each task if the system’s dynamics is learned well.

Real-time localization is an important mission for self-driving cars and it is difficult to achieve precise pose information in dynamic environments. In this paper, a novel localization method is proposed to estimate the pose of self-driving cars using a 3D-LiDAR sensor. First, the multi-frame curb features and laser intensity features are extracted. Meanwhile, based on the high-precision curb map generated offline, obstacles on road are detected using region segmentation methods and their features are removed. Furthermore, a map-matching method is proposed to match the features to the map, a robust iterative closest point algorithm is utilized to deal with curb features along with a probability search method dealing with intensity features. Finally, two separate Kalman filters are used to fuse the low-cost global positioning systems and map-matching results. Both offline and online experiments are carried out in dynamic environments and the results demonstrate the accuracy and robustness of the proposed method.

This study is concerned with the tracking control problem for nonlinear uncertain robotic systems in the presence of unknown actuator nonlinearities. A novel adaptive sliding controller is designed based on a robust disturbance observer without any prior knowledge of actuator nonlinearities and system dynamics. The proposed control strategy can guarantee that the tracking error eventually converges to an arbitrarily small neighborhood of zero. Simulation results are included to demonstrate the effectiveness and superiority of the proposed strategy.

This paper presents a new algorithm for lidar data assimilation relying on a new forward model. Current mapping algorithms suffer from multiple shortcomings, which can be related to the lack of clear forward model. In order to address these issues, we provide a mathematical framework where we show how the use of coarse model parameters results in a new data assimilation problem. Understanding this new problem proves essential to derive sound inference algorithms. We introduce a model parameter specifically tailored for lidar data assimilation, which closely relates to the local mean free path. Using this new model parameter, we derive its associated forward model and we provide the resulting mapping algorithm. We further discuss how our proposed algorithm relates to usual occupancy grid mapping. Finally, we present an example with real lidar measurements.

As an important part of the manufacturing industry, redundant robots can undertake heavy and tough tasks, which human operators are difficult to sustain. Such onerous and repetitive industrial manipulations, that is, positioning and carrying, impose heavy burdens on the load bearing for redundancy robots’ joints. Under the circumstances of long-term and intense industrial operations, joints of redundant robots are conceivably to fall into functional failure, which may possibly cause abrupt joint lock or freeze at unknown time instants. Therefore, task accuracy by end-effectors tends to diminish considerably and gradually because of broken-down joints. In this paper, a sparsity-based method for fault-tolerant motion planning of redundant robots is provided for the first time. The developed fault-tolerant redundancy resolution approach is defined as L1-norm based optimization with immediate variables involved to avoid discontinuity in the dynamic solution process. Meanwhile, those potential faulty joint(s) can be located by the designed fault observer with the proposed fault-diagnosis algorithm. The proposed fault-tolerant motion planning method with fault diagnosis is dynamically optimized by resultant primal dual neural networks with provable convergence. Moreover, the sparsity of joint actuation by the proposed method can be enhanced by around 43.87% and 36.51%, respectively, for tracking circle and square paths. Simulation and experimental findings on a redundant robot (KUKA iiwa) prove the efficacy of the developed defect tolerant approach based on sparsity.

Welding is a complex manufacturing process. Its quality depends on the welder skills, especially in welding complex paths. For consistency in modern industries, the arm robot is used to accomplish this task. However, its programming and reprogramming are time consuming and costly and need an expert programmer. These limit the use of robots in medium and small industries. This paper introduces a new supervised learning technique for programming a 4-degree of freedom (DOF) welding arm robot with an automatic feeding electrode. This technique is based on grasping the welding path control points and motion behavior of an expert welder. This is achieved by letting the welder move the robot end effector, which represents the welding torch, through the welding path. At the path control points, the position and speed are recorded using a vision system. Later, these data are retrieved by the robot to replicate the welding path. Several 2D paths are tested to assess the proposed approach accuracy and programming time and easiness in comparison with the common one. The results prove that the proposed approach includes fewer steps and consumes less programming time. Moreover, programming can be accomplished by the welder and no need for an expert programmer. These enhancements will improve the share of robots in welding and similar industries.

As an important index to quantitatively measure the motion performance of a manipulator, motion reliability is affected by many factors, such as joint clearance. The present research utilized a UR10 manipulator as the research object. A factor mapping model for influencing the motion reliability was established. The link flexibility factor, joint flexibility factor, joint clearance factor, and Denavit–Hartenberg (DH) parameters were comprehensively considered in this model. The coupling relationship among the various factors was concisely expressed. Subsequently, the nonlinear response surface method was used to calculate the reliability and sensitivity of the manipulator, which provided an applicable reference for its trajectory planning and motion control. In addition, a data-driven fault diagnosis method based on the kernel principal component analysis (KPCA) was used to verify the motion accuracy and sensitivity of the manipulator, and joint rotation failure was considered as an example to verify the accuracy of the KPCA method. This study on the motion reliability of the manipulator is of great significance for the current motion performance, adjusting the control strategy and optimizing the completion effect of the motion task of a manipulator.

This study presents a novel context awareness multihuman–robot interaction (MHRI) system that allows multiple operators to interact with a robot. In the system, a monocular multihuman 3D pose estimator is first developed with the convolutional neural network. The estimator first regresses a set of 2D joints representations of body parts and then restores the 3D joints positions based on these 2D representations. Further, the 3D joints are assigned to the corresponding individual with a priority–redundancy association algorithm. The whole 3D pose of each person is reconstructed in real time, even in crowded scenes containing both self-occlusion of the body and inter-person occlusion. Then, the identities of multiple persons are recognized with action context and 3D skeleton tracking to improve interactive efficiency. For context-awareness multitask interaction, the robot control strategy is designed based on target goal generation and correction. The generated goal is taken as a reference to the model predictive controller (MPC) to generate motion trajectory. Different interactive requirements are adapted by adjusting the weight parameters of the energy function of the MPC controller. Multihuman–robot interactive experiments, including dynamic obstacle avoidance (human–robot safety) and cooperative handling, demonstrate the feasibility and effectiveness of the MHRI, and the safety and collaborative efficiency of the system are evaluated with HRI metrics.

Success of any needle-based medical procedures depends on accurate placement of the needle at the target location. However, accurate targeting and control of flexible self-actuating (active) needle are challenging. We have developed a shape memory alloy-actuated flexible needle steered by a 3D Cartesian robot and performed a comparative study of four, non-model-based, coordinated control methodologies for the combined robot steering and flexible-needle insertion process for surgical interventions. Investigated four controllers are: proportional–integral–derivative (PID), PID with the cubic of positional error term (PID-P3), static PID sliding mode controller, and robust adaptive PID sliding mode controller (RAPID-SMC). Relative efficacies of these controllers are demonstrated by performing experiements using a tissue-mimicking soft material phantom. Results from experiments have reavealed that RAPID-SMC is superior to other three controllers.

Robotic systems have shown capabilities to perform inspection tasks in dangerous and difficult-to-access environments, such as those found in different components of power plants. However, most of the current robotic inspection technology is designed for specific components. Aerial robots, commonly termed as Drones, have raised an option to inspect a wider range of structural components. Nevertheless, current aerial inspecting technology still relies on a human pilot with limited line of sight, field of view and a reduced perception as the drone flies away, which prevents performing close-quarter inspection in intricate, structurally complex and GPS-denied environments. This work introduces offline inspection path generation methods based on robotics-integrated to manufacturing techniques. One method uses computer-aided manufacturing (CAM) techniques and the other an additive manufacturing (AM) approach to generate the flight path. That is to say, the drone would fly along the path described by a 3D (Three Dimensional) printer’s extruder or a CNC (Computer Numerical Control) machining tool, enabling to fly very close the structure even in physical structures with a complex geometry. Once the trajectories are generated, they are introduced for its validation in the Gazebo robotics simulator. Simulation results demonstrate the proper performance of the method and confirm that this approach can be used for close inspection of structuralcomponents.

In this paper, a new method is proposed to find a feasible energy-efficient path between an initial point and goal point on uneven terrain and then to optimally traverse the path. The path is planned by integrating the geometric features of the uneven terrain and the biped robot dynamics. This integrated information of biped dynamics and associated cost (energy) for moving toward the goal point is used to define the value of a new speed function at each point on the discretized surface of the terrain. The value is stored as a matrix called the dynamic transport cost (DTC). The path is obtained by solving the Eikonal equation numerically by fast marching method (FMM) on an orthogonal grid, by using the information stored in the DTC matrix. One step of walk on uneven terrain is characterized by 10 footstep parameters (FSPs); these FSPs represent the position of swinging foot at the starting and ending time of the walk, orientation, and state (left or right) of support foot. A walking dataset was created for different walking conditions (FSPs), which the biped robot is likely to encounter when it has to walk on the uneven terrain. The corresponding energy optimal hip and foot trajectory parameters (HFTPs) are obtained by optimization using a genetic algorithm (GA). The created walk dataset is generalized by training a feedforward neural network (NN) using the scaled conjugate gradient (SCG) algorithm. The Foot placement planner gives a sequence of foot positions and orientations along the obtained path, which is followed by the biped robot by generating real-time optimal foot and hip trajectories using the learned NN. Simulation results on different types of uneven terrains validate the proposed method.

This paper presents a trajectory planning method based on multi-objective optimization, including time optimal and jerk optimal for the manipulators in the presence of obstacles. The proposed method generates a trajectory configuration in the joint space with kinematic and obstacle constraints using quintic B-spline. Gilbert–Johnson–Keerthi detecting algorithm is utilized to detect whether there is a collision and obtain the minimum distance between the manipulator and obstacles. The degree of constraint violations is introduced to redefine the Pareto domination, and the constrained multi-objective particle swarm algorithm (CMOPSO) is adopted to solve the time-jerk optimization problem. Finally, the Z-type fuzzy membership function is proposed to select the best optimal solution in the Pareto front obtained by CMOPSO. Test results show the effectiveness of the proposed method.

In this paper, we propose a set of robust training methods for deep reinforcement learning to transfer learning acquired in one control task to a set of previously unseen control tasks. We improve generalization in commonly used transfer learning benchmarks by a novel sample elimination technique, early stopping, and maximum entropy adversarial reinforcement learning. To generate robust policies, we use sample elimination during training via a method we call strict clipping. We apply early stopping, a method previously used in supervised learning, to deep reinforcement learning. Subsequently, we introduce maximum entropy adversarial reinforcement learning to increase the domain randomization during training for a better target task performance. Finally, we evaluate the robustness of these methods compared to previous work on simulated robots in target environments where the gravity, the morphology of the robot, and the tangential friction coefficient of the environment are altered.

Stroke has a considerable incidence in the world population and would cause sequelae in the upper limbs. One way to increase the efficiency in the rehabilitation process of patients with these sequelae is through robot-assisted therapy. The present study developed a portable robotic orthosis called Pinotti Portable Robotic Exoskeleton (PPRE) and validated its functioning in clinical tests. The static and dynamic parts of the device modules are described. Design issues, such as heavyweight and engine positioning, have been optimized. The implementation of control was through a smartphone application that communicates with a microcontroller to perform desired movements. Four individuals with motor impairment of the upper limbs due to stroke performed clinical tests to validate the device. Participants did not mention pain, discomfort, tingling, and paresthesia. The robotic device showed the ability to perform the flexion and extension movements of the fingers and elbow. The PPRE was confirmed to be adequate and functional at different levels of motor impairment assessed. The orthosis presented advantages over the currently existing devices, concerning its biomechanical functioning, portability, comfort, and versatility. Thus, the apparatus has the great innovative potential to become a device for home use, serving as an aid to the therapist and facilitating the rehabilitation of patients after an injury. In a larger sample, future studies are needed to assess the effect of a robotic orthosis on the level of rehabilitation in individuals with upper limb impairment.

This paper presents the design and simulation of a Parallel Kinematic (PK) testbed for head impacts. The proposed design is presented as a novel head impact testbed using a parallel platform as main motion simulation mechanism. The testbed is used to give a motion to a head mannequin to impact against a steel plate. In addition, the platform in the testbed allows to modify the orientation of the head mannequin model to evaluate different types of impacts. The testbed has been modeled with software MS ADAMS® to evaluate its performance with a dynamic simulation and to characterize the testbed design during top and lateral impact events. Results show that PK testbed gives a significant force and acceleration to the head mannequin at the moment of the impact.

In this paper, several physical activity-based human–computer interaction (HCI) games which are developed and implemented for the improvement of attention, emotion, and sensory–motor coordination will be presented. The interface and the difficulty levels of these games are specially designed for the use of people with different age groups and disabilities. The games involve physical activities for the fulfillment of some basic HCI tasks which require hand and arm motion for control, such as fruit picking and air hockey, with adaptive difficulty levels based on varying parameters of the games and human performance. In the fruit picking game, several fruit images are moving from top to the bottom of the screen. Objective is to collect apples while avoiding the pears. The player’s hand will control the basket that collects the fruits. In the air hockey game, the player will try to score goals against computer-controlled opponent. The player’s hand will control the paddle to hit the puck to score or to defend his/her goal area. The player’s hand is recognized by Kinect RGB-D sensors in both games. Aim of the adaptive difficulty-based system is keeping the players engaged in the games. The games are tested with a group of deaf children (3.5–5 years) as a part of an ongoing project,1 to decrease the stress of the children and increase their positive emotions, attention, and sensory–motor coordination before the audiology tests. The game performances and the evaluation of the therapists show that the games have a positive impact on the children. The games are also tested with a group of adults as a control group, where a mobile EEG device is employed to detect the attention levels. For this purpose, the adults also attended a third game featuring a maze and controlled with Myo sensors.

This article proposes a robust and adaptive controller for industrial robot arms with multiple degrees of freedom without the need for velocity measurement. Many of the controllers designed for manipulators are model-based and require detailed knowledge of the system model. In contrast to these methods, this paper proposes a model-free controller using the Philips q-Bernstein operator as universal approximator. The designed controller can approximate uncertainties including external disturbances and unmodeled dynamics based on its universal approximation capability. Besides, most of the controllers revealed for robot arms are torque-based, which is not a realistic presumption from a practical point of view. In the proposed control method, the voltage applied to the actuator is considered as the control signal. However, unlike many voltage-based methods, the need to know the exact models of the system and the actuator has been eliminated in the presented method. Also, adaptive rules are extracted during the Lyapunov analysis to ensure system stability. Finally, to analyze the performance of the presented controller, this method is simulated for an industrial robot arm, and the results are analyzed. The proposed methodology is also compared to those of a strong state-of-the-art approximator, the Chebyshev neural network.

A kinetostatic approach applied to the design of a backflip strategy for quadruped robots is proposed in this paper. Inspired by legged animals and taking the advantage of the leg workspace, this strategy provides an optimal design idea for the low-cost quadruped robots to achieve self-recovery after overturning. Through kinetostatic and energy analysis, a four-stepped backflip strategy based on the selected rotation axis with minimum energy is proposed, with a process of selection, lifting, rotating, and protection. The kinematic factors that affect the backflip are investigated, along with the relationship between the design parameters of the leg and trunk being analyzed. At the end of this paper, the strategy is validated by a simulation and experiments with a prototype called DRbot, demonstrating that the strategy endows the robot a strong self-recovery ability in various terrains.

A comprehensive dynamic modeling and actuator torque minimization of a new symmetrical three-degree-of-freedom (3-DOF) 3-PṞR spherical parallel manipulator (SPM) is presented. Three actuating systems, each of which composed of an electromotor, a gearbox and a double Rzeppa-type driveshaft, produce input torques of the manipulator. Kinematics of the 3-PṞR SPM was recently studied by the author (Zarkandi, Proc. Inst. Mech. Eng. Part C J. Mech. Eng. Sci. 2020, https://doi.org/10.1177%2F0954406220938806). In this paper, a closed-form dynamic equation of the manipulator is derived with the Newton–Euler approach. Then, an optimization problem with kinematic and dynamic constraints is presented to minimize torques of the actuators for implementing a given task. The results are also verified by the SimMechanics model of the manipulator.

This article introduces a robust adaptive controller–observer structure for robotic manipulators such that the need for joints speed measurement is removed. Besides, it is presumed that the system model has uncertainties and is subject to disturbances, and the proposed method must eliminate the impact of these factors on the system response. According to this, for the first time in the robotics field, a model-free scheme is developed based on the Bernstein–Stancu polynomial. The universal approximation property of the Bernstein–Stancu polynomial enables it to accurately estimate the lumped uncertainty, including unmodeled dynamics and disturbances. Moreover, to increase the efficiency of the controller–observer structure, adaptive rules have been proposed to update polynomial coefficients. The boundedness of all system errors is proven using the Lyapunov theorem. Finally, the proposed robust Adaptive controller–observer is applied on a planer robot, and the results are precisely analyzed. The results of the proposed approach are also compared with two state-of-art powerful approximation methods.