Grasping force control is important for multi-fingered robotic hands to stabilize the grasped object. Humans are able to adjust their grasping force and react quickly to instabilities through tactile sensing. However, grasping force control through tactile sensing with robotic hands is still relatively unexplored. In this paper, we make use of tactile sensing for multi-fingered robot hands to adjust the grasping force to stabilize unknown objects without prior knowledge of their shape or physical properties. In particular, an online detection module based on Deep Neural Network (DNN) is designed to detect contact events and object material simultaneously from tactile data. In addition, a force estimation method based on Gaussian Mixture Model (GMM) is proposed to compute the contact information (i.e., contact force and contact location) from tactile data. According to the results of tactile sensing, an object stabilization controller is then employed for a robotic hand to adjust the contact configuration for object stabilization. The spatio-temporal property of tactile data is exploited during tactile sensing. Finally, the effectiveness of the proposed framework is evaluated in a real-world experiment with a five-fingered Shadow Dexterous Hand equipped with BioTac sensors.

Learning skills autonomously is a particularly important ability for an autonomous robot. A promising approach is reinforcement learning (RL) where agents learn policy through interaction with its environment. One problem of RL algorithm is how to tradeoff the exploration and exploitation. Moreover, multiple tasks also make a great challenge to robot learning. In this paper, to enhance the performance of RL, a novel learning framework integrating RL with knowledge transfer is proposed. Three basic components are included: 1) probability policy reuse; 2) dynamic model learning; and 3) model-based Q-learning. In this framework, the prelearned skills are leveraged for policy reuse and dynamic learning. In model-based Q-learning, the Gaussian process regression is used to approximate the Q-value function so as to suit for robot control. The prior knowledge retrieved from knowledge transfer is integrated into the model-based Q-learning to reduce the needed learning time. Finally, a human-robot handover experiment is performed to evaluate the learning performance of this learning framework. Experiment results show that fewer exploration is needed to obtain a high expected reward, due to the prior knowledge obtained from knowledge transfer.

In‐hand manipulation is a fundamental ability for multi‐fingered robotic hands that interact with their environments. Owing to the high dimensionality of robotic hands and intermittent contact dynamics, effectively programming a robotic hand for in‐hand manipulations is still a challenging problem. To address this challenge, this work employs deep reinforcement learning (DRL) algorithm to learn in‐hand manipulations for multi‐fingered robotic hands. A reward‐shaping method is proposed to assist the learning of in‐hand manipulation. The synergy of robotic hand postures is analysed to build a low‐dimensional hand posture space. Two additional rewards are designed based on both the analysis of hand synergies and its learning history. The two additional rewards cooperating with an extrinsic reward are used to assist the in‐hand manipulation learning. Three value functions are trained jointly with respect to their reward functions. Then they cooperate to optimise a control policy for in‐hand manipulation. The reward shaping not only improves the exploration efficiency of the DRL algorithm but also provides a way to incorporate domain knowledge. The performance of the proposed learning method is evaluated with object rotation tasks. Experimental results demonstrated that the proposed learning method enables multi‐fingered robotic hands to learn in‐hand manipulation effectively.

Contact-rich manipulation tasks are difficult to program to be performed by robots. Traditional compliance control methods, such as impedance control, rely excessively on environmental models and are ineffective in the face of increasingly complex contact tasks. Reinforcement learning (RL) has now achieved great success in the fields of games and robotics. Autonomous learning of manipulation skills can empower robots with autonomous decision-making capabilities. To this end, this work introduces a novel learning framework that combines deep RL (DRL) and variable impedance control (VIC) to achieve robotic massage tasks. A skill policy is learned in joint space, which outputs the desired impedance gain and angle for each joint. To address the limitations of the sparse reward of DRL, an intrinsic curiosity module (ICM) was designed, which generates the intrinsic reward to encourage robots to explore more effectively. Simulation and real experiments were performed to verify the effectiveness of the proposed method. Our experiments demonstrate that contact-rich massage skills can be learned through the VIC–DRL framework based on the joint space in a simulation environment, and that the ICM can improve learning efficiency and overall performance in the task. Moreover, the generated policies have been demonstrated to still perform effectively on a real-world robot.

Autonomous robotic massage holds the potential to alleviate the workload of nurses and improve the quality of healthcare. However, the complexity of the task and the dynamic of the environment present significant challenges for robotic massage. This paper presents a vision-based robotic massage (VBRM) framework that facilitates autonomous robot massaging of the human body while ensuring safe operation in a dynamic environment. The VBRM framework allows the operator to define the massage trajectory by drawing a 2D curve on an RGB image. An interactive trajectory planning method is developed to calculate a 3D massage trajectory from the 2D trajectory. This method accounts for potential movements of the human body and updates the planned trajectory using rigid point cloud registration. Additionally, a hybrid motion/force controller is employed to regulate the motion of the robot’s end-effector, considering the possibility of excessive contact force. The proposed framework enables the operator to adjust the massage trajectory and speed according to their requirements. Real-world experiments are conducted to evaluate the efficacy of the proposed approach. The results demonstrate that the framework enables successful planning and execution of the massage task in a dynamic environment. Furthermore, the operator has the flexibility to set the massage trajectory, speed, and contact force arbitrarily, thereby enhancing human–machine interaction.

Summary: Determining an appropriate grasp configuration for multi-finger grasping is difficult due to the complexity of robotic hands. The multi-finger grasp planning should consider not only geometry constraints of objects but also kinematics and dynamics of robotic hand. In this paper, a precision grasp-planning framework is presented for multi-finger hand to grasp unknown objects. First, the manipulation capabilities of the robotic hand are analyzed. The analysis results are further used as bases for the precision grasp planning. Second, the superquadric (SQ) fitting method is used for approximating unknown object models. Finally, a local–global optimization method is implemented to find appropriate grasp configurations for dexterous hand. The presented planning framework is validated in simulation experiments. Simulation results demonstrated that the presented grasp-planning framework enables the multi-finger hand to grasp unknown objects effectively.