Legged animals can adapt to complex terrains because they can step or jump over obstacles. Their application of foot force is determined according to the estimation of the height of an obstacle; then, the trajectory of the legs is controlled to clear the obstacle. In this paper, we designed a three-DoF one-legged robot. A spring-loaded inverted pendulum model was employed to control the jumping. Herein, the jumping height was mapped to the foot force by mimicking the jumping control mechanisms of animals. The foot trajectory in the air was planned using the Bézier curve. Finally, the experiments of the one-legged robot jumping over multiple obstacles of different heights were implemented in the PyBullet simulation environment. The simulation results demonstrate the effectiveness of the method proposed in this paper.

The appeal of quadruped robots lies not only in their ability to mimic the diverse and agile locomotion of animals but also in their potential to integrate with human artistic expression to achieve complex multi-skill movements. Continuous multi-skill motion in quadruped robots requires the realization of diverse, continuous, and long-horizon behaviors, involving a broader state space and more complex motion generation and transitions. This presents significant challenges, including sparse rewards, incomplete data, long-horizon motion planning, and the design of fine-grained motion transitions. In this work, we categorize quadruped robot skills into three types: rhythmic motions, expressive motions, and high-dynamic motions and generate reference trajectories for each category using central pattern generators, animation design, and motion capture, respectively. We then design an asymmetric neural network architecture and employ an imitation–reinforcement learning algorithm to train policies for generating these three types of motions. By composing multiple motion skill trajectories, we avoid long-horizon motion planning; by leveraging reinforcement learning, we enable smooth and continuous skill transitions; and by introducing a two-stage reference state initialization curriculum, the robot is able to switch from arbitrary states to the target motion skill. Moreover, during training, the policy imitates only key characteristics of the reference motions rather than strictly tracking fixed trajectories, making it more robust. Finally, we achieve robust motion skill generation and seamless transitions on a quadruped robot equipped with a 6-DoF manipulator, validating the effectiveness and feasibility of the proposed multi-skill generation and transition method.

Canine animals excel at running and jumping, exhibiting remarkable agility and making them popular bionic models for legged robots. In this study, we utilized motion capture equipment to record a Malinois dog jumping over hurdles and circular holes while running, obtaining motion trajectory data of the trunk, limbs, head, and tail. Subsequently, a digital model of both the Malinois and the environment was constructed, and the limb motion data, as well as the spatial relationships between the dog and the obstacles, were analyzed in both temporal and spatial dimensions. From this analysis, we summarized the behavioral strategies and kinematic patterns underlying the Malinois’ running and obstacle‐crossing process. The main findings are as follows: (1) The obstacle‐crossing process follows a specific footfall sequence and can be divided into three major phases: takeoff phase, flight phase, and landing phase. (2) During the takeoff phase, the pitch angle of the trunk at liftoff exhibits an arctangent function relationship with the height and distance of the obstacle, while the liftoff velocity is determined by the obstacle height and distance. (3) In the flight phase, the head and tail movements contribute to adjusting trunk posture. (4) During the landing phase, the forelimbs touch the ground first, and the virtual leg formed by the hip and foot generates a spring‐like effect upon contact. Finally, we applied data retargeting and optimization methods to reproduce the Malinois’ obstacle‐crossing behavior on a quadruped robot, verifying the practicality of our data and research findings. Additionally, we proposed a performance evaluation method for animals and legged robots in jumping over obstacles while running.