Journal of Robotics

Considering the complexity of multidimension parameters and the mechanical performance of a 6-DOF robotic crusher, a multiobjective optimization function based on the transmission index and condition number is established. As an important operation in the screw theory, the reciprocal product between the transmission wrench screw of an actuator and the output twist screw of the mantle assembly is used to represent the instantaneous power. The expression of transmission index is derived according to the principle that constraint wrench screws apply no work to the mantle assembly. It can be used as a criterion to evaluate the transmission performance. Then, based on the Jacobian matrix, the equation of condition number is constructed which provides a criterion for evaluating kinematic accuracy. Finally, the workspace and singularity of the 6-DOF robotic crusher are analyzed to verify the rationality of the optimized variables. The results show that the optimized structure can completely crush the material in the workspace and effectively avoid singularity, which provides a basis for practical application.

This paper studies the design of pneumatic soft-bodied bionic basic execution unit with soft-rigid combination, which can be used as an actuator for pneumatic soft-bodied robots and soft-bodied manipulators. This study is inspired by structural characteristics and motion mechanism of biological muscles, combined with the nonlinear hyperelasticity of silica gel and the insertion of thin leaf spring structure in the nonretractable layer. Response surface analysis and numerical simulation algorithm are used to determine the optimal combination of structural dimension parameters by taking the maximum output bending angle of the basic executing unit as the optimization objective. Based on Odgen’s third-order constitutive model, the deformation analysis model of the basic execution unit is established. The physical model of pneumatic soft-bodied bionic basic execution unit is prepared through 3D printing, shape deposition, soft lithography, and other processing methods. Finally, the motion and dynamic characteristics of the physical model are tested through experiments and result analysis, thus obtaining curves and empirical formulas describing the motion and dynamic characteristics of the basic execution unit. The relevant errors are compared with the deformation analysis model of the execution unit to verify the feasibility and effectiveness of the design of the pneumatic soft-bodied bionic basic execution unit. The above research methods, research process, and results can provide a reference for the research and implementation of pneumatic and hydraulic driven soft-bodied robots and grasping actuators of soft-bodied manipulators.

The inherent highly nonlinear coupling and system uncertainties make the controller design for a flexible-joint robot extremely difficult. The goal of the control of any robotic system is to achieve high bandwidth, high accuracy of trajectory tracking, and high robustness, whereby the high bandwidth for flexible-joint robot is the most challenging issue. This paper is dedicated to design such a link position controller with high bandwidth based on sliding-mode technique. Then, two control approaches ((1) extended-regular-form approach and (2) the cascaded control structure based on the sliding-mode estimator approach) are presented for the link position tracking control of flexible-joint robot, considering the dynamics of AC-motors in robot joints, and compared with the singular perturbation approach. These two-link position controllers are tested and verified by the simulation studies with different reference trajectories and under different joint stiffness.

Vanilla policy gradient methods suffer from high variance, leading to unstable policies during training, where the policy’s performance fluctuates drastically between iterations. To address this issue, we analyze the policy optimization process of the navigation method based on deep reinforcement learning (DRL) that uses asynchronous gradient descent for optimization. A variant navigation (asynchronous proximal policy optimization navigation, appoNav) is presented that can guarantee the policy monotonic improvement during the process of policy optimization. Our experiments are tested in DeepMind Lab, and the experimental results show that the artificial agents with appoNav perform better than the compared algorithm.

Posture-position control is the fundamental technology among multilegged robots as it is hard to get an effective control on rough terrain. These robots need to constantly adjust the position-posture of its body to move stalely and flexibly. However, the actual footholds of the robot constantly changing cause serious errors during the position-posture control process because their foot-ends are basically in nonpoint contact with the ground. Therefore, a position-posture control algorithm for multilegged robots based on kinematic correction is proposed in this paper. Position-posture adjustment is divided into two independent motion processes: robot body position adjustment and posture adjustment. First, for the two separate adjustment processes, the positions of the footholds relative to the body are obtained and their positions relative to the body get through motion synthesis. Then, according to the modified inverse kinematics solution, the joint angles of the robot are worked out. Unlike the traditional complex closed-loop position-posture control of the robot, the algorithm proposed in this paper can achieve the purpose of reducing errors in the position-posture adjustment process of the leg-foot robot through a simple and general kinematic modification. Finally, this method is applied in the motion control of a bionic hexapod robot platform with a hemispherical foot-end. A comparison experiment of linear position-posture change on the flat ground shows that this method can reduce the attitude errors, especially the heading error reduced by 55.46%.

Wall defect detection is an important function for autonomous decoration robots. Object detection methods based on deep neural networks require a large number of images with the handcrafted bounding box for training. Nonetheless, building large datasets manually is impractical, which is time-consuming and labor-intensive. In this work, we solve this issue to propose the low-shot wall defect detection algorithm using deep reinforcement learning (DRL) for autonomous decoration robots. Our algorithm first utilizes the attention proposal network (APN) to generate attention regions and applies AlexNet to extract the features of attention patches to further reduce computation. Finally, we train our method with deep reinforcement learning to learn the optimal detection policy. The experiments are implemented on a low-shot dataset in which images are collected from real decoration environments, and the experimental results show the proposed method can achieve fast convergence and learn the optimal detection policy for wall defect images.