A control system for controlling a legged robot comprises a processor and a memory. The control system initializes, in response to receiving a task, a probabilistic filter with parameters associated with a state of a reference trajectory of the legged robot, wherein the parameters are predetermined for the task and encode the reference trajectory including a combination of different reference trajectories for coordinated motion primitives of different actuators of the legged robot moving the legged robot according to the task, decodes the parameters to generate the reference trajectory, and executes, in response to receiving a feedback signal, the probabilistic filter to iteratively track the state of the reference trajectory satisfying a performance objective with respect to the state of the legged robot to update the parameters.

A robot that is mobile and includes: a controller that controls the robot; and an action unit that performs an action based on an instruction from the controller. The controller causes the action unit to perform a preliminary action before the robot starts to move in a place where the robot and a person are present together.

A robotic dog empowered by generative artificial intelligence (Gen-AI) is disclosed, capable of autonomously performing essential tasks such as guiding visually impaired individuals, detecting drugs and arms, and providing companionship. The robotic dog's lifelike design includes a head, eyes, ears, a nose, a mouth with teeth, a neck, a body, four legs with paws, and a tail, all meticulously crafted to mimic the appearance and behavior of a real dog. A trained AI model functions as the brain, processing environmental data captured by video cameras, audio microphones, and sensors to provide guidance commands to a control system that control the movements of the robotic dog. A well-trained live dog can serve as a teacher for one or multiple robotic dogs using a generative AI-based real-time training method, enabling efficient and effective training of robotic dogs.

A wheel-legged robot is considered as n.sup.th-order inverted pendulum model including a wheel, n links, and n revolute joints. First links are from at least two leg mechanisms of the wheel-legged robot. The wheel is from mobile wheels respectively connected to the at least two leg mechanisms. A balance control method for the robot includes: obtaining an actual state vector of the wheel-legged robot at a first moment; calculating an equivalent state vector at the first moment based on the actual state vector at the first moment; establishing a sliding surface based on the equivalent state vector at the first moment; determining a force and torque instruction for whole-body joints based on the sliding surface, the equivalent state vector at the first moment, and a dynamics equation of the wheel-legged robot; and separately controlling the n revolute joints at a second moment according to the force and torque instruction.

A wireless charging method for an intelligent quadruped robot includes guiding the quadruped robot to the charging panel, detecting a surface condition of the charging panel via the detection device, generating a first signal once a foreign object is present on the charging panel, otherwise, generating a second signal, controlling the charging power supply in a turned-off state when in idle; remaining the charging power supply in the turned-off state based on the first signal; switching from the turned-off state to a turned-on state based on the second signal; and charging a rechargeable battery in the quadruped robot. The method achieves autonomous movement to reduce manual intervention, and prevents charging anomalies or hazards to ensure charging safety.

An intelligent security method, system, and storage medium based on a quadruped robot are provided. The method includes receiving first coordinates corresponding to a target object and formulating a navigation path; dynamically correcting the navigation path during movement; when a distance between the quadruped robot and the target object reaches a preset tracking threshold, activating a surveillance camera and capturing a contour image of the target object; calculating a similarity between the contour image and a preset reference image; when the similarity is less than the preset value, initiating an attack and/or an alarm action, and calculating second coordinates; updating the navigation path and continuously tracking. Problems of limited surveillance coverage of traditional systems, inability to actively track moving targets, and untimely response to unexpected targets are solved.

Systems and methods for semantic robot hazard avoidance with multi-modal prompting are provided. In one aspect, a method includes receiving a user input indicative of one or more hazards in an environment of the robot and image data indicative of the one or more hazards in the environment. The method also includes generating one or more segments of the image data. Each of the one or more segments corresponds to at least one of the one or more hazards indicated by the user input. The method further includes identifying a semantic label for each of the one or more segments, generating a hazard map including a location of each of the one or more segments and the corresponding semantic label, and navigating the robot through the environment based at least in part on the hazard map.

A method in accordance with at least some embodiments of the present technology includes determining first hazard information about a human in an environment at a first time. The method further includes decelerating a mobile robot in the environment based at least partially on the first hazard information. The method further includes determining second hazard information about the human at a second time after the first time. The method further includes reconfiguring the mobile robot based at least partially on the second hazard information. Reconfiguring the mobile robot includes moving the mobile robot from a standing configuration to a non-standing configuration. The method further includes determining third hazard information about the human at a third time after the second time. Finally, the method includes causing a safe operating stop of the mobile robot based at least partially on the third hazard information.

According to one embodiment, a method, computer system, and computer program product for resonance avoidance is provided. The embodiment may include identifying one or more entities traversing a structure. The embodiment may also include identifying movement information and characteristic information for the one or more entities. The embodiment may further include performing a digital twin simulation of the structure based on the identified movement information and characteristic information. The embodiment may also include, in response to determining resonance between a natural frequency of the structure and a movement pattern of each entity based on the digital twin simulation, identifying one or more movement changes for each entity to eliminate resonance of the structure.

Method, apparatuses and computer program products for training machine learning models for logistics robots and routing the robots are disclosed. A method of training an ML model involves providing a route to a logistics robot, obtaining route issue indications from the robot when it fails to traverse the route as expected, and associating map objects with the issue locations. The trained model can then be used to determine the likelihood of route issues for a specific logistics robot type based on the presence of certain map objects along the route. The disclosure further involves calculating route penalty value(s) based on the likelihood and updating the route accordingly, including selection of a logistics robot type for the route.