Load capacity is an important index to reflect the practicability of legged robots. Existing research into quadruped robots has not analyzed their load performance in terms of their structural design and control method from a systematic point of view. This paper proposes a structural design method and crawling pattern generator for a planar quadruped robot that can realize high-payload locomotion. First, the functions required to evaluate the leg’s load capacity are established, and quantitative comparative analyses of the candidates are performed to select the leg structure with the best load capacity. We also propose a highly integrated design method for a driver module to improve the robot’s load capacity. Second, in order to realize stable load locomotion, a novel crawling pattern generator based on trunk swaying is proposed which can realize lateral center of mass (CoM) movement by adjusting the leg lengths on both sides to change the CoM projection in the trunk width direction. Finally, loaded crawling simulations and experiments performed with our self-developed quadruped robot show that stable crawling with load ratios exceeding 66% can be realized, thus verifying the effectiveness and superiority of the proposed method.

The jumping motion of legged robots is an effective way to overcome obstacles in the rugged microgravity planetary exploration environment. At the same time, a quadruped robot with a manipulator can achieve operational tasks during movement, which is more practical. However, the additional manipulator will restrict the jumping ability of the quadruped robot due to the increase in the weight of the system, and more active degrees of freedom will increase the control complexity. To improve the jumping height of a quadruped robot with a manipulator, a bio-inspired take-off maneuver based on the coordination of upper and lower limbs is proposed in this paper. The kinetic energy and potential energy of the system are increased by driving the manipulator-end (ME) to swing upward, and the torso driven by the legs will delay reaching the required peak speed due to the additional load caused by the accelerated ME. When the acceleration of ME is less than zero, it will pull the body upward, which reduces the peak power of the leg joints. Therefore, the jumping ability of the system is improved. To realize continuous and stable jumping, a control framework based on whole-body control was established, in which the quadruped robot with a manipulator was a simplified floating seven-link model, and the hierarchical optimization was used to solve the target joint torques. This method greatly simplifies the dynamic model and is convenient for calculation. Finally, the jumping simulations in different gravity environments and a 15° slope were performed. The jump heights have all been improved after adding the arm swing, which verified the superiority of the bio-inspired take-off maneuver proposed in this paper. Furthermore, the stability of the jumping control method was testified by the continuous and stable jumping.

A quadruped robot with a manipulator that combines dynamic motion and manipulation capabilities will greatly expand its application scenarios. However, the addition of the manipulator raises the center of mass of the quadruped robot, increasing complexity in motion control and posing new challenges for maintaining balance on sloped terrains. To address this, a balance control method based on whole-body synergy is proposed in this study, emphasizing adaptive adjustment of the robot system’s overall balance through effective utilization of the manipulator’s active motion. By establishing a mapping relationship between the manipulator and the robot’s attitude angle under system equilibrium, the desired manipulator motion is guided by real-time estimates of terrain angles during motion, enhancing motion efficiency while ensuring robot balance. Furthermore, to enhance motion tracking accuracy, the optimization of system angular momentum and manipulator manipulability is incorporated into hierarchical optimization tasks, improving manipulator controllability and overall system performance. Simulation and experimental results demonstrate that the quadruped robot with a manipulator exhibits reduced velocity and attitude angle fluctuations, as well as smoother foot-end force dynamics during climbing motions with the addition of manipulator adaptive adjustment. These results validate the effectiveness and superiority of the manipulator-based adaptive adjustment strategy proposed in this paper.

Aiming at highly dynamic locomotion and impact mitigation, this paper proposes the design and implementation of a symmetric legged robot. Based on the analysis of the three-leg topology in terms of force sensitivity, force production, and impact mitigation, the symmetric leg was designed and equipped with a high torque density actuator, which was assembled by a custom motor and two-stage planetary. Under the kinematic and dynamic constraints of the robot system, a nonlinear optimization for high jumping and impact mitigation is proposed with consideration of the peak impact force at landing. Finally, experiments revealed that the robot achieved a jump height of 1.8 m with a robust landing, and the height was equal to approximately three times the leg length.

Accurate 6D pose estimation is critical for robots to perform grasping tasks successfully. In recent years, numerous 6D pose estimation algorithms have been widely applied in robotic grasping experiments. However, in complex environments, industrial robots often struggle to extract effective image features, leading to inaccurate 6D pose estimations and subsequent grasping failures. Additionally, improper grasping sequences caused by stacked parts increase the risk of collisions, further elevating the probability of grasping failure. To overcome these challenges, we presented CF6D, a collision-free grasping network based on 6D pose estimation. Specifically, we devised a keypoint prediction network that utilizes a least squares fitting algorithm to exactly detect object keypoints and estimate 6D poses. The network also outputs an optimal grasping sequence, enabling the robot to effectively grasp stacked parts based on precise poses and a reasonable order. On the relevant dataset, our method obviously better than the existing methods in terms of accuracy and robustness. Furthermore, test results on a real-world robotic platform demonstrate that this method excels in practical grasping tasks, meeting the demands of industrial applications.