Power management methods and power systems for smart devices

Abstract

We disclose herein a power management system for a smart device, including a power source, a monitoring unit, a management unit, and a power controller. The power source is electrically coupled to the smart device. The monitoring unit is configured to collect power source data of the power source and behavioral data of at least one actuator of the smart device. The management unit receives the power source data and behavioral data and is communicatively isolated from external communication networks and the smart device. The power controller can terminate the electrical connection between the power source and the smart device. During operation, the management unit sends instructions to the power controller to terminate the electrical connection between the power source and the smart device based on abnormalities in the power source data or the behavioral data.

Description

151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00POWER MANAGEMENT METHODS AND POWER SYSTEMS FOR SMART DEVICESCROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 749,718, filed January 27, 2025, and titled "POWER MANAGEMENT METHOD AND POWER SYSTEM FOR SMART DEVICES,” which is incorporated herein by reference in its entirety7.TECHNICAL FIELD

[0002] The present technology generally relates to power management methods and power systems for smart devices, including associated devices, systems, and methods.BACKGROUND

[0003] The rapid iteration of artificial intelligence (Al) and robotics technologies has spurred the large-scale deployment of smart devices capable of environmental perception, autonomous decision-making, and task execution. These devices are continuously penetrating diverse scenarios in human production and daily life, covering core areas such as home services, healthcare, industrial manufacturing, public services, and transportation hubs. However, as the performance of smart devices continues to improve and their interaction with humans becomes increasingly sophisticated, operational security7has become a key bottleneck restricting the industry’s large-scale development.

[0004] Existing power source solutions for smart devices (e.g., devices with Al integrated) focus on four main dimensions: meeting the device’s power source needs, reducing the overall size and weight of the device, improving energy' efficiency, and achieving scalability' of the power source architecture. Regarding the core security issues of artificial intelligence systems, traditional solutions primarily rely on software control strategies and built-in security functions within the device for protection. In this type of design, the smart device controller continuously monitors its own operational behavior. If a security risk is detected, it w ill automatically attempt to reduce the device’s operating load or directly initiate a proactive system shutdown process.

[0005] However, because smart devices possess autonomous learning and decisionmaking capabilities, entrusting all security functions to the Al system will reveal significant151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00weaknesses in extreme scenarios. When the Al system malfunctions, experiences configuration errors, or exposes software vulnerabilities, it often fails to promptly identify dangerous behaviors or quickly shut down the system. More seriously, if the Al system suffers external intrusion, hacker attacks, or malicious tampering, it can directly trigger abnormal or dangerous actions in the smart device, such as suddenly charging into crowds, unreliably operating heavy equipment, or activating various tools and instruments in dangerous modes.

[0006] Al-powered drones and unmanned aerial vehicles also face similar risks. In the event of software malfunctions or hacking attacks, these drones may stray into no-fly zones, fly uncontrollably near crowds, or even collide with buildings and infrastructure. Traditional flight control systems typically integrate safety logic with mission logic into a single software architecture, while most drone propulsion systems lack an independent safety layer (i.e., they cannot directly cut off power to the propulsion motors and critical loads when abnormal behavior is detected).

[0007] Furthermore, traditional power source systems typically cannot independently perceive actuator behavior, cannot monitor the spatial relationship between robots or drones and surrounding people, and cannot apply scene-appropriate safety limits based on the presence of infants, children, or vulnerable groups. At the same time, these systems generally lack dedicated mechanisms to identify attempts by one robot to interfere with or disable another robot’s safety system. In multi-robot or robot-drone hybrid scenarios, even if tampering is detected, a coordinated safety response mechanism is often lacking.

[0008] Additionally, a safety system cannot reliably determine whether the Al control system has modified its own programs, models, or control strategies solely through software interfaces. Robots or Al systems may autonomously change their operational behavior without notifying external hardware, even exhibiting an ‘“awakened” state or unpredictable actions. Therefore, if a safe power source system relies solely on explicit program change notifications from the Al control system, it cannot serve as the sole means of safety protection.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Many aspects of the present disclosure can be better understood with reference to the following drawings. The components in the drawings are not necessarily to scale. Instead, emphasis is placed on illustrating clearly the principles of the present disclosure. The drawings151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00should not be taken to limit the disclosure to the specific embodiments shown, but are provided for explanation and understanding.

[0010] Figure 1 is a block diagram illustrating a power management system configured in accordance with various embodiments of the present technology.

[0011] Figures 2 and 3 are block diagrams illustrating the power management system configured in accordance with various embodiments of the present technology.

[0012] Figure 4 is a block diagram illustrating a computer system configured in accordance with various embodiments of the present technology.

[0013] Figure 5 is a flow diagram of a method of managing a power source to a smart device in accordance with various embodiments of the present technology.DETAILED DESCRIPTIONI. Introduction

[0014] The present technology is generally directed to a power management system and a power management method for smart devices (e.g., intelligent devices, Al-integrated devices, Al devices). The power management system physically connects to the power source path of the smart device, operates independently of the controller of the smart device and external network at the logic and communication levels, and can autonomously determine whether an abnormal or unsafe condition exists based on power source data, actuator behavior data, and / or image data from an independent visual sensor. When an unsafe or unsecured condition is determined, the system can limit current, voltage, power, and / or cut off all power to the smart device through the power controller, thus serving as a last line of defense for safety.

[0015] The power management system and method can be widely applied to fields such as, for example, ground robots, industrial robotic arms, autonomous vehicles, and artificial intelligence drones. Specifically, it can provide dedicated power management methods and safe power systems for various artificial intelligence (Al) systems, such as, for example, humanoid robots, functional robots, autonomous vehicles, artificial intelligence drones, and industrial robotic arms.

[0016] In some embodiments, the power management system can directly control the power source path of smart devices, detect abnormal power sources and actuator behavior, monitor the environment in real time via independent visual sensors, and autonomously limit or151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00cut off power when unsafe conditions are detected. The safety trigger mechanism can identify observable behaviors of smart devices, regardless of whether these behaviors stem from device malfunctions, external hacking attacks, or changes in autonomous control behavior. Furthermore, the power management system can integrate machine learning technology to dynamically adjust safety thresholds by learning the devices’ daily operating patterns, while ensuring that the protective capabilities of the Al system are not weakened. The system can also support controlled manual intervention via a protected physical interface, be equipped with factory -preset limits and daily automatic reset functions, and explicitly prohibit the use of weapons — unless temporary authorization from a human operator is obtained.

[0017] In some embodiments, the power management system includes a power source, a monitoring unit, at least one visual sensor, an independent management unit, and a power controller. The power source can be used to provide power to the power management system of the smart device. In some embodiments, the power source is selected from at least one of: a battery system, a supercapacitor system, a fuel cell system, or other DC power sources suitable for smart devices. The monitoring unit can collect power source data (e.g., current, voltage, power) from the smart device and obtain behavioral data (e.g., torque, speed, position, direction of rotation, joint angle, velocity, acceleration) from one or more actuators or motors of the smart device. In some embodiments, the monitoring unit can include current sensors, voltage sensors, shunt resistors, Hall effect sensors, torque sensors, encoders, joint angle sensors, and / or other sensor interfaces. These sensor interfaces are directly electrically connected to the management unit, without requiring intermediaries through the control system of the smart device. The visual sensor can be used to monitor the workspace surrounding the smart device and detect external objects, including infants, children, adults, and animals. The power management system is equipped with at least one visual sensor. The image data collected by the sensor is directly transmitted to the management unit, while the smart device’s control system cannot directly access this data. In some embodiments, the visual sensor(s) are selected from a single camera, multiple cameras, a depth camera, a stereo camera, an infrared camera, and / or other imaging devices. Additionally, the power controller is the sole connection between the smart device power management system and the smart device. The power controller, acting as the sole connector, is connected in series between the power source and the smart device load. The power controller can include actuators, sensors, and / or control electronics. In some embodiments, the power controller can be implemented using semiconductor switches, solid-state relays, contactors, DC / DC converters, and / or combinations thereof, and has the functions of limiting151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00current, limiting voltage, limiting power, and / or completely disconnecting the power source from the smart device.

[0018] In some embodiments, the management unit includes an image analysis module for classifying external objects and implementing specific category safety thresholds and distance-based safety zones. The management unit generates and stores a safety model based on power source data and actuator behavior data under normal operating conditions, and receives data in real time during operation. During operation, the management unit compares the realtime data with the safety model to determine if any abnormal or unsafe conditions exist. When such a condition is detected, it sends a command to the power controller to limit or cut off the power source. The management unit can be implemented using a microcontroller, microprocessor, system-on-a-chip, programmable logic device, and / or a combination thereof. The management unit can run dedicated security firmware and software, and cannot be programmed by the controller of the smart device. The management unit further stores the safety model and related safety thresholds.

[0019] In other embodiments, the management unit includes a machine learning module. This module stores time-series logs of power source data, actuator behavior data, and image data collected during daily operation, and uses these logs to train and update the safety model. The machine learning module can automatically adjust safety’ thresholds to reflect learned normal behavior patterns while inhibiting, or even preventing, thresholds from exceeding preset safety hardware limits. Confirmed safety events and near misses can be used as labeled data to optimize the safety' model and reduce false alarm rates. Since the adaptive program of the control system of the smart device cannot be directly detectable, the power management system only monitors observable behavior and considers situations that continuously deviate from learned normal behavior as potential safety hazards, regardless of whether the cause is hardware failure, hacker attack, or system modification. The machine learning module updates safety thresholds by learning the daily behavior patterns of the devices. In some embodiments, a manual configuration interface with an external touch panel with a protective cover allows for temporary adjustment of thresholds and manual weapon usage permissions. In further embodiments, the system has factory-preset safety limits and supports automatic reset at night.

[0020] In some embodiments, the power source, power controller, and / or management unit are integrated into a power control box, the outer surface of which is printed with a visible safety' label. A visual sensor can identify this safety label and detect any robotic arm or tool151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00attempting to approach, touch, and / or damage the power control box. Upon detecting tampering, the management unit instructs the power controller to cut off power to the smart device. In multirobot or hybrid robot / drone environments, one smart device can monitor the power control boxes of other smart devices and trigger a collaborative shutdown mechanism, including shutting down its own device, upon detecting tampering.

[0021] In some embodiments, the powder management system is equipped with a manual configuration interface, consisting of an external touch panel covered by a mechanical cover or door panel. When an authorized person opens the cover, the user can select regular operation files and multiple special task configuration files via the touch panel. The authorized person can temporarily adjust the safety thresholds for special tasks, such as thresholds for handling heavy loads or high-speed operation. Factory-preset safety thresholds and modes are preconfigured through factory’ training programs on smart devices or reference devices. User modifications to the thresholds are considered temporary overwrite settings, and the system automatically restores the factory7default thresholds at a predetermined time (e.g., during a nighttime reset) for continued use the following day.

[0022] In some embodiments, the power management system includes a dangerous object (e.g., weapons, oversized objects) control module. The management unit acquires image data through a visual sensor, identifies weapons such as knives, firearms, clubs, or baseball bats, and marks them as prohibited items. By default, the system prohibits smart devices from accessing or using these weapons. An external touch panel provides a manual w eapon usage authorization function, allowing authorized personnel to temporarily grant access to specific w eapon items under strict time, period, and security restrictions. All temporary activation of weapon-related permissions can be automatically revoked when the system resets at night.

[0023] In some embodiments, the power management system constructs a safety model using power source data and actuator behavior data during normal operation of the smart device. In actual operation, the system collects power source data, actuator behavior data, and image data in real time, and updates the safety model and safety thresholds through machine learning technology. The system can automatically detect abnormal or unsafe states, including abnormal actuator operation, insufficient personnel safety distance, tampering with the power control box, and weapon-related anomalies, and autonomously implement power limiting or power cut-off operations through the powder controller. This safety logic operates solely based on observable151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00behavior and does not require any program change notifications from the control system of the smart device.

[0024] In some embodiments, a method of operating the power management system can include: (i) collecting power source data and actuator behavior data under normal operating conditions; (ii) generating a safety model based on the collected data: (iii) real-time acquisition of pow er source data, actuator behavior data, and image data during operation; (iv) updating the safety model and threshold settings using the machine learning module; (v) determining whether there is an abnormal or unsafe state by comparing the safety model with the safety model and thresholds; and (iv) limiting or cutting off the power source to the smart device automatically through the power controller when an unsafe condition is detected. In some embodiments, the method can further include: (i) receiving temporary threshold adjustments for special tasks can be received via an external touch panel; (ii) detecting tampering with the power management system; (iii) identifying weapon-related items; (iv) implementing weapon disabling policies; (v) identifying abnormal patterns that indicate malfunctions, attacks, or self-modifying control behavior, logging event data; and (vi) resetting thresholds to factory' defaults every night.

[0025] In some embodiments, the system includes a non-volatile computer-readable medium storing instructions that, when executed by a management unit, can trigger the secure power management method. This storage medium can be implemented using read-only memory’, flash memory, or other persistent memory within the system.

[0026] The power management system can be applied in scenarios such as: (i) if a smart robot malfunctions and starts walking on its own w hile charging at night, the power management system can detect the anomaly and cut off the pow er to the robot to inhibit harm to family members or property; (ii) if a robot goes out of control and attacks people, the power management system can detect the anomaly and cut off the power to the robot to reduce the risk of injury; (iii) if a self-driving car goes out of control, the power management system can detect the anomaly and slow dowvi or stop the vehicle by limiting current or cutting off power. Another extreme scenario is when the Al system receives an external intrusion and is controlled to exhibit abnormal behavior, such as when an autonomous vehicle is hacked and controlled to cause extreme harm.

[0027] In some aspects of the present technology, the power management system is completely disconnected from the Internet and has no communication connection with the Al system (smart device). When the Al goes out of control, it can independently' carry' out accurate151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00and timely processing to ensure the safety of humans and the environment. Further, the power management system can serve as the last line of defense for Al devices. The system is not connected to artificial intelligence or any external network. The pre-embedded safety initial model in the system has functions such as data collection and statistical analysis to summarize the normal power consumption behavior model of artificial intelligence devices. When the artificial intelligence device exhibits abnormal behavior, the system can detect the abnormality through model comparison and issue an early warning. In special circumstances, it can independently manage power without being controlled by artificial intelligence, thereby ensuring safety7. Further, this safety logic is based entirely on observable or detectable behavior and does not rely on program change notifications from the control system of the smart device. Any behavior that continuously deviates from the learned normal patterns of power, motion, and / or environmental context is considered a potential hazard, regardless of whether it stems from system failure, external hacking, and / or self-regulating control behavior. This system is suitable for humanoid robots, functional robots, autonomous vehicles, Al drones, and other smart devices, providing a final layer of safety for these devices.

[0028] Certain details are set forth in the following description and in Figures 1-5 to provide a thorough understanding of various embodiments of the present technology. In other instances, well-known structures, materials, operations, and / or systems often associated with Al, power management, and / or the like are not shown or described in detail in the following disclosure to avoid unnecessarily obscuring the description of the various embodiments of the technology. Those of ordinary skill in the art will recognize, however, that the present technology can be practiced without one or more of the details set forth herein, and / or with other structures, methods, components, and so forth. Moreover, although many of the devices and systems are described herein in the context of client-server computing, the present technology7can be used in other computing systems, such as cloud computing, distributed computing, cluster computing, personal computing, mobile computing, and / or the like.

[0029] The terminology used below is to be interpreted in its broadest reasonable manner, even though it is being used in conjunction with a detailed descnption of certain examples of embodiments of the technology. Indeed, certain terms may even be emphasized below; however, any terminology7intended to be interpreted in any restricted manner will be overtly and specifically defined as such in this Detailed Description section.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00

[0030] The accompanying Figures depict embodiments of the present technology and are not intended to be limiting of its scope unless expressly indicated. The sizes of various depicted elements are not necessarily drawn to scale, and these various elements may be enlarged to improve legibility. Component details may be abstracted in the Figures to exclude details such as the position of components and certain precise connections between such components when such details are unnecessary for a complete understanding of how to make and use the present technology. Many of the details, dimensions, angles, and other features shown in the Figures are merely illustrative of particular embodiments of the disclosure. Accordingly, other embodiments can have other details, dimensions, angles, and features without departing from the present technology. In addition, those of ordinary' skill in the art will appreciate that further embodiments of the present technology can be practiced without several of the details described below.

[0031] In the Figures, identical reference numbers identify identical, or at least generally similar, elements. To facilitate the discussion of any particular element, the most significant digit or digits of any reference number refer to the Figure in which that element is first introduced. For example, the power management system WO is first introduced and discussed with reference to Figure 1.

[0032] The headings provided herein are for convenience only and should not be construed as limiting the subject matter disclosed. To the extent any materials incorporated herein by reference conflict with the present disclosure, the present disclosure controls.II. Selected Embodiments of Power Management Systems

[0033] Figure 1 is a block diagram illustrating a power management system 100 (“system 100”) configured in accordance with various embodiments of the present technology'. The system 100 includes a power source 111 for powering the smart device 200. a monitoring unit 101, at least one visual sensor 102 (individually labeled a first visual sensor 102a and a second visual sensor 102b), a management unit 112, and a power controller 113 positioned in a power path 114 between the power management system 100 and the smart device 200.

[0034] In the illustrated embodiment, the power management system 100 includes a subsystem power control box 110 that includes a management unit 112, a power source 111, and a power controller 113. The power source 111 delivers power to the management unit 112 and the power controller 113, which directs power to the smart device 200 along the power path 114. The management unit 112 is independent of the smart device 200 and does not have a signal connection with the smart device 200. The management unit 112 is not connected to any external151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00communication network. The only connection between the power management system 100 and the smart device 200 is achieved through the power path 114 controlled by the power controller 113.

[0035] In some embodiments, an outer surface of the power control box 110 features visual safety markings, such as special patterns, symbols, or codes recognizable by visual sensors and image analysis modules (e.g., the visual sensors can be specially configured and oriented to monitor the power control box 110 and its visual safety markings in real time). The power control box 110 can be physically installed on or near the smart device 200.

[0036] In the illustrated embodiment, the first visual sensor 102a and the second visual sensor 102b collect visual information of the environment and transmit this information directly to the management unit 112. At the same time, the smart device 200 feeds back its own behavior data and power source data to the monitoring unit 101, which then transmits this data to the management unit 112. The management unit 112 receives information from the visual sensors 102 and the monitoring unit 101, acquiring the power status of the smart device 200. The management unit 112 sends control commands to the power controller 113, which is connected to the smart device 200, that controls the pow er flow along the power path 114 from the power source 111. This can include current limiting, voltage limiting, and / or power cut-off from the power source 111 to the smart device 200. By controlling the power source 111, the management unit 112 manages the power of the smart device 200.

[0037] In the illustrated embodiment, the management unit 112 is used to perform the following steps: acquire power source data, process the power source data to form a power source data model (e.g., safety model), acquire real-time power source data, compare and analyze the real-time power source data with the power source data model, determine whether there is any abnormality in the real-time power source data, and if an abnormality is determined, send a control power source output command to power controller 113.

[0038] In some embodiments, the monitoring unit 101 collects power source data under different working conditions, including parameters such as current, voltage, and power. These operating conditions cover various tasks, loads, speeds, and / or environmental conditions. In addition, the system collects actuator behavior data, such as torque, speed, position, rotation direction, joint angles, and / or acceleration. For example, for drones equipped with artificial intelligence, actuator behavior data can include propeller motor cunent, speed, thrust level, and / or attitude control signals. These datasets form the basis for building a safety model, as151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00described in greater detail below with reference to Figure 5. The management unit 112 compares and analyzes real-time data with the actuator behavior characteristics stored in the safety model. When a deviation from the safety model exceeds a preset threshold, the management unit 112 detects an abnormal or unsafe condition and can terminate power to the smart device 200.

[0039] In some embodiments, the power management system 100 features a manual configuration interface, which is mounted as an external touch panel on the power control box 110. The touch panel can be covered by a mechanical protective cover or door panel that includes locking or latching functions. The touch panel can only be operated when the protective cover is open, effectively reducing the risk of accidental or unauthorized configuration changes. Authorized operators can select general operation files and one or more special task profiles via the touch panel. The authorized operators can temporarily adjust safety thresholds within preset ranges, such as allowing smart devices to carry additional heavy loads under higher permissible torque and current limits, achieving higher speeds and accelerations, or operating within temporarily extended operating ranges. All adjustments made via the touch panel can be stored as temporarily overridden settings associated with the selected special task mode.

[0040] In some embodiments, factor}7safety thresholds and configuration files are preset by the manufacturer through factory training programs. These programs can involve controlled operation of smart or reference devices through representative tasks, recording power source and actuator behavior data to define safety ranges. These factor}7thresholds constitute the factory default safety settings.

[0041] In some embodiments, the management unit 112 imposes constraints on each specific task profile, such as a maximum allowed duration, a maximum allowed number of loops, or required physical confirmation input on the touch panel. It also sets global safety limits for thresholds to ensure adjustments do not exceed hardware safety limits. At preset times (such as during a nighttime reset), the management unit 112 can automatically clear all temporarily overridden settings and reset all safety thresholds and profiles to factory defaults, preparing them for standard use the following day.

[0042] Figures 2 and 3 are block diagrams illustrating the power management system 100 configured in accordance with various embodiments of the present technology. As shown in Figure 2 the monitoring unit 101 can be connected between the power source 111 and the power controller 113; or, as shown in Figure 3, the monitoring unit 101 can be connected between the power controller 113 and the smart device 200. The power controller 113 controls the power151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00source output of the power source 111 according to a control command received from the management unit 112.

[0043] Figure 4 is a block diagram illustrating a computer system 400 configured in accordance with various embodiments of the present technology. At least some operations of the system 100 described herein can be implemented on the computer system 400. The computer system 400 can include one or more central processing units (“processors”) 402, main memory 406, non-volatile memory 410, video displays 418, input / output devices 420, control devices 422 (e.g., keyboard and pointing devices), drive units 424 including a storage medium 426, and a signal generation device 430 that are communicatively connected to a bus 416. The bus 416 is illustrated as an abstraction that represents one or more physical buses and / or point-to-point connections that are connected by appropriate bridges, adapters, or controllers. The bus 416, therefore, can include a system bus, a Peripheral Component Interconnect (PCI) bus or PCI-Express bus, a HyperTransport or industry standard architecture (ISA) bus, a small computer system interface (SCSI) bus, a universal serial bus (USB), IIC (I2C) bus, an Institute of Electrical and Electronics Engineers (IEEE) standard bus 416 (also referred to as “Firewire”), and / or the like.

[0044] The computer system 400 can share a similar computer processor architecture as that of a desktop computer, tablet computer, personal digital assistant (PDA), mobile phone, game console, music player, wearable electronic device (e.g., a watch or fitness tracker), network-connected (“smart”) device (e g., a television or home assistant device), virtual / augmented reality systems (e.g., a head-mounted display), or another electronic device capable of executing a set of instructions (sequential or otherw ise) that specify action(s) to be taken by the computer system 400.

[0045] While the main memory 406, non-volatile memory 410, and storage medium 426 (also called a “machine-readable medium”) are shown to be a single medium, the term “machine-readable medium” and “storage medium” should be taken to include a single medium or multiple media (e.g., a centralized / distributed database and / or associated caches and servers) that store one or more sets of instructions 428. The term “machine-readable medium” and “storage medium” shall also be taken to include any medium that is capable of storing, encoding, or carrying a set of instructions for execution by the computer system 400. In some embodiments, the non-volatile memory 410 or the storage medium 426 is a non-transitory, computer-readable151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00storage medium storing computer instructions, which can be executed by the one or more central processing units (‘“processors”) 402 to perform functions of the embodiments disclosed herein.

[0046] In general, the routines executed to implement the embodiments of the disclosure can be implemented as part of an operating system or a specific application, component, program, object, module, or sequence of instructions (collectively referred to as “computer programs”). The computer programs typically include one or more instructions (e.g., instructions 404, 408, 428) set at various times in various memory and storage devices in a computer device. When read and executed by the one or more processors 402, the instruction(s) cause the computer system 400 to perform operations to execute elements involving the various aspects of the disclosure.

[0047] Moreover, while embodiments have been described in the context of fully functioning computer devices, those skilled in the art will appreciate that the various embodiments are capable of being distributed as a program product in a variety of forms. The disclosure applies regardless of the particular type of machine or computer-readable media used to actually effect the distribution.

[0048] Further examples of machine-readable storage media, machine-readable media, or computer-readable media include recordable-type media such as volatile and non-volatile memory 410, floppy and other removable disks, hard disk drives, optical discs (e.g., Compact Disc Read-Only Memory (CD-ROMS), Digital Versatile Discs (DVDs)), and transmission-type media such as digital and analog communication links.

[0049] The techniques introduced here can be implemented by programmable circuitry (e.g.. one or more microprocessors), software and / or firmware, special-purpose hardwired (i.e., non-programmable) circuitry, or a combination of such forms. Special-purpose circuitry can be in the form of one or more application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), and / or the like.III. Selected Embodiments of Power Management Methods

[0050] Figure 5 is a flow diagram of a method 500 of managing a power source (e.g., power source 111) to a smart device (e.g., smart device 200) in accordance with various embodiments of the present technology. At block 502, the method 500 can include obtaining baseline power source data of the smart device when the power system (e.g., system 100) is operating normally. The power source data includes current, voltage, and / or power. Typically, a151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00large amount of power source data of the smart device during normal operation can be obtained through targeted laboratory experiments and / or simulated data collection of hypothetical application scenarios. By monitoring the power source status of the power system, the power consumption of the smart device can be fully understood. In other embodiments, the power source data of the smart device under different working conditions is obtained as the input data of the model. The power source data and the corresponding different working conditions are used as labels to generate a training set.

[0051] At block 504, the method 500 can include establishing a power source data model, also referred to as a safety model, through training with a large amount of the baseline power source data. In some embodiments, the power source data model is a power consumption data model of the smart device during normal operation. In other embodiments, the model continuously collects power source data, accumulates data over a long period of time, and / or actively corrects the power source data model.

[0052] In some embodiments, blocks 502 and 504 can include: (i) acquiring real-time power source data of the smart device under different operating conditions as input data of the model, (ii) using the real-time power source data and the corresponding different operating conditions as labels to generate a training set; (iii) using the training set to train and update the model parameters until the power source data model is generated.

[0053] In some embodiments, the safety model can include statistical characteristic curves, limit curves, time series templates, and a training model describing normal power and actuator behavior. This model can be generated by training a machine learning module, whose training set contains training data with power source data and actuator behavior data as input and corresponding operating conditions as labels. The machine learning module can employ algorithms such as regression models, neural networks, cluster analysis, and anomaly detection, and / or use other pattern recognition techniques suitable for modeling normal behavior.

[0054] During operation, a management unit (e.g., management unit 112) continuously records the real-time power source data, actuator behavior data, and / or image data. The machine learning module analyzes this log data to continuously optimize and update the safety model. The system labels two types of events: (i) confirmed safety events, such as shutdowns or power outages triggered by exceeding safety thresholds; and (ii) "near-miss" events that approach the safety threshold but do not result in a shutdown. This data is used for model improvement, enabling the machine learning module to automatically adjust the safety thresholds — reducing151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00false alarms while maintaining or even improving the sensitivity to detecting real dangerous (e.g., hazardous, unsafe) conditions.

[0055] In some embodiments, the management unit sets maximum safety limits for all thresholds based on hardware performance and risk analysis. If the machine learning module malfunctions or fails an internal consistency check, the management unit can automatically switch to a conservative preset safety threshold or initiate a safe shutdown procedure.

[0056] At block 506, the method 500 can include monitoring the real-time power source data at the power source end. Specifically, this includes comparing the real-time power source data with the threshold or waveform limits or waveforms generated in the power source data model. In some aspects of the present technology, by monitoring the power source status, the power consumption of the smart device can be obtained, enabling the power system to monitor and control powder consumption independently of the smart device. If the smart device malfunctions and becomes uncontrollable, the power system, independent of the smart device's sensing and algorithms, detects the abnormal power consumption data and actively controls the power source.

[0057] At block 508, the method 500 can include comparing and analyzing the real-time power source data with the power source data model. In some embodiments, this includes statistically analyzing and regressing the power source data model during power consumption. During normal operation, by analyzing or training real-time power source data from various scenarios, the power source data model is upgraded and updated, making it more accurate. In some embodiments, the real-time power data and power source data mode can be used to identify abnormal behavior patterns (e.g., abnormal state), such as internal faults in the control system of the smart device, external hacker attacks, or self-modify ing control behaviors.

[0058] At block 510, the method 500 can include determining whether there are any abnormalities in the real-time power source data. Specifically, this can include determining if the real-time power source data exceeds the threshold or waveform limits or if there are abnormal waveforms. In some embodiments, the thresholds or waveform limits can be provided by the power source data model or determined by setting a certain proportion based on the pow er source data model. In other embodiments, temporary7threshold adjustments for special tasks can be received via an external touch panel. If the real-time power source data exceeds these limits, an anomaly is determined. If an abnormality is found, the method 500 can advance to block 512, which can include automatically issuing an early warning and / or automatically controlling the151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00power source output, thereby ensuring the safety of the smart device and the surrounding environment in extreme situations. Controlling the power source output includes current limiting, voltage limiting, power limiting, or power cutoff. Issuing an early warning primarily involves the power system sending a warning signal to the smart device. If the smart device does not take timely control measures, the power system can directly control the power source output, forming a final safety line. That is, once the power source data model detects abnormal power consumption behavior (such as sudden abnormal changes in current or power, or abnormal waveforms changing over time), the power management system autonomously issues an early warning, intervenes, or cuts off power. After the power management system autonomously controls the power source output, it can be manually restored. At this time, the power source data model can be updated and adjusted according to the manual restoration scenario, completing the update of the power source data model and making the power source data model more accurate. If an abnormality is not found, the method 500 can return to block 506 and repeat.

[0059] In some embodiments, the method 500 can additionally and / or alternatively include: (i) detecting tampering with the system controller, identifying dangerous objects (e.g., weapons), and enforcing weapon disabling policies; and (ii) identifying abnormal patterns that indicate malfunctions, attacks, or self-modifying control behavior, logging event data, and resetting thresholds to factory defaults every' night.

[0060] In some aspects of the present technology, when the smart device malfunctions and becomes uncontrollable, the power system, independent of the smart device’s sensing and algorithms, detects the abnormal power consumption data and thus actively controls the power source (e g., current limiting, voltage limiting, power limiting, power cut-off) to ensure the safety of the smart device and the surrounding environment.

[0061] In some embodiments, the system collects actuator behavior data, such as torque, speed, position, rotation directionjoint angles, and / or acceleration. For drones equipped with artificial intelligence, actuator behavior data can include propeller motor current, speed, thrust level, and / or attitude control signals. These datasets form the basis for building the safety model. The management unit compares and analyzes this data with the actuator behavior characteristics stored in the safety model. When the deviation exceeds a preset threshold, the management unit detects an abnormal or unsafe condition.

[0062] In some embodiments, abnormal situations include: sudden torque surges, sustained torque saturation, violent shaking exceeding the allowable range, movement in151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00prohibited directions, repeated collisions or impacts, and / or abnormal thrust patterns in the drone that can indicate loss of control. Once these abnormal signals are detected, the management unit can immediately instruct the power controller to take countermeasures such as current limiting, torque reduction, deceleration, thrust reduction, and / or complete power cut-off of the smart device.

[0063] In some aspects of the present technology, because the actuator behavior data is collected independently of the control system of the smart device, even if the controller of the smart device issues unsafe commands or malfunctions, the safety power management system can still detect abnormal actuator behavior and intervene by limiting or cutting off power.

[0064] In some embodiments, visual sensors (e.g., visual sensors 102) collect image data of the workspace surrounding the smart device. The management unit is equipped with an image analysis module that processes this image data to identify external objects and entities, including infants, toddlers, children, adults, and animals. The image analysis module can categorize external objects into different classes, such as category one for infants or toddlers, category two for children, and category three for adults. Each category corresponds to a different safety threshold. When near infants or toddlers, the maximum permissible speed, acceleration, torque, and / or contact force can be significantly lower than in the adult zone. The management unit can also define distance-based safety zones, such as a moderately restricted outer zone and a strictly restricted or completely shut-down inner zone when an object is detected.

[0065] During operation, the management unit can first calculate the distance and relative orientation between the robot’s end effector and the target. Based on the target type and distance parameters, the system calculates safe thresholds for contact force, torque, speed, and / or power. If real-time actuator data shows that these limits have been exceeded or are about to be exceeded, the management unit can immediately instruct the power controller to reduce output power and, if necessary, initiate an emergency shutdown procedure. This logic also applies to drones equipped with artificial intelligence — when approaching or exceeding altitude and lateral distance restrictions for people and buildings, the system automatically reduces thrust or performs a controlled landing to ensure safety.

[0066] In some aspects of the present technology, image analysis is implemented within the management unit and is logically isolated from the smart device. The smart device cannot lower safety limits based on its own perception or planning. Therefore, the independent visual151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00safety layer provides an additional layer of protection for vulnerable groups such as infants and young children.

[0067] In some embodiments, when a robotic arm, gripper, or tool of the smart device approaches a power control box (e.g., the power control box 110), the management unit detects that a visual safety marker has entered a preset area captured by the sensor. If the marker appears continuously within a specific distance range for more than a preset duration, or appears simultaneously with a gripper / tool performing actions such as grabbing, releasing, loosening, cutting, disassembling, and / or removing the power control box, the system can determine that tampering has occurred.

[0068] When tampering is detected, the management unit instructs the power controller (e.g., power controller 113) to cut off power to actuators and / or control electronics, thereby severing the entire power source to the smart device. In some aspects of the present technology', the pow er management system effectively inhibits robots (whether or not they have been hacked) from disabling or bypassing their own power control box without authorization. The management unit can also record tampering events and send alerts to human monitoring personnel.

[0069] In a multi-robot collaborative environment, multiple smart devices can each be equipped with an independent powder management system. In practice, the visual sensors of the first smart device can monitor the power control box status of the second smart device in real time. When tampering is detected in the power control box of the second smart device, the management unit of the first smart device can immediately trigger the shutdown operation of the second device or switch it to a safe low-power mode. Simultaneously, the first smart device can also proactively shut down its own power, thereby effectively reducing potential risks.

[0070] In some embodiments, this environment can include a hybrid configuration of ground robots and Al-driven drones. For example, Al-driven drones can carry downward or sideways vision sensors to monitor power control boxes mounted on the ground robots. When a drone detects unauthorized operations, it sends a signal to the safety system or monitoring unit of the affected ground robot, initiating an emergency shutdown procedure. Conversely, ground robots equipped with upward vision sensors can monitor abnormal proximity of flying drones to people or buildings in real time and trigger collaborative safety response mechanisms.

[0071] In another embodiment, when any device in the preset smart device group detects tampering or combined tampering with weapon status, a broadcast or wired triggering device151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00can put all devices in the group into a safe state and stop their movement. For drones, they can be guided to a safe area for controlled landing or hovering to await human intervention.

[0072] In some embodiments, the power management system has a weapon-disabling function. The image analysis module in the management unit collects image data through a visual sensor to detect and classify objects within the workspace of the smart device. Specific objects such as knives, guns, clubs, and baseball bats are classified as dangerous objects (e.g., weapons, oversized objects). These dangerous objects are marked as disabled objects by default.

[0073] The management unit can detect when the end effector or gripper of the smart device is grasping or about to grasp a weapon-like object. The system can also recognize weapon-using behavior when the device moves the weapon-like object by waving, striking, or pointing it at a person. If such behavior is detected without permission for weapon use, the management unit can immediately instruct the power controller to reduce the permissible speed and torque to a safe minimum or to cut off (e g., eliminate) all power to the device. This event can be logged and archived for subsequent analysis.

[0074] In some embodiments, the touch panel can include a manual weapon use authorization function. Authorized operators can explicitly authorize the smart device to operate items of a specified weapon category within a limited time and under controlled conditions (such as maintenance, testing, or special safety tasks). Each manual weapon use authorization comes with constraints, including the maximum allowed duration, the allowed number of operations or cycles, and a task description entered by the operator. During the authorized period, the system can enforce stricter safety thresholds than usual to ensure weapon operation safety.

[0075] In some embodiments, the management unit records each weapon use permit and its associated sensor data and events in non-volatile memory. When the permit period or usage cycle expires, a security incident occurs, or a nighttime reset is performed, the management unit automatically revokes the weapon use permit and restores the system to its default state — at which point all weapon categories are prohibited from use. In some aspects of the present technology, the smart device cannot bypass the weapon-disabling function, nor can it extend weapon use permits beyond the limits set by the management unit.

[0076] In some circumstances, the power management system can operate in a fail-safe mode. When the management unit detects an internal fault, interruption of critical sensor input, inconsistency in the safety model, and / or an anomaly in the machine learning module, it can automatically switch the system to the default safe state. At this time, the system can take safety151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00measures such as reducing power limits, limiting speed and torque, and / or cutting off all power to smart devices.

[0077] In the event of a downtime or security incident, the management unit can store power source data logs, actuator behavior data, image data, and / or relevant information on violation thresholds or rules in tamper-proof storage before and after the incident. These logs support subsequent investigations and continuous improvement of security policies. In some embodiments, these logs are stored in a tamper-proof memory.

[0078] Power-limiting or power-off commands issued by the management unit take precedence over any commands from the smart device. The smart device cannot prevent or override power-off operations initiated by the management unit. In some embodiments, an emergency stop input port can be connected to the management unit such that when the emergency stop button is triggered, the system can immediately perform a power-off operation and generate a corresponding log record.

[0079] In some circumstances, the control system of the smart device can malfunction, be attacked by external hackers, or modify its behavior or control strategies in ways not disclosed to external hardware. In such cases, the power management system cannot reliably detect when the internal program changes itself. Instead, the power management system treats the control system of the smart device as a black box, focusing only on the observable behavior of the smart device.

[0080] In some embodiments, the machine learning module and the management unit together constitute an abnormal behavior detection system. For example, when a device continuously deviates from the normal pattern of learned power source data, actuator behavior data, or image data, it can be considered a potential safety hazard even if a hard safety limit is not immediately triggered. For example, if the smart device begins to adopt new motion patterns, new action sequences, or new environmental interaction strategies that were not present during training, the management unit can switch to a more conservative operating mode, reducing risk by limiting speed, torque, and power, until sufficient experience is accumulated to determine whether the new pattern is safe.

[0081] Generally, the power management system does not pre-determine or differentiate the causes of abnormal behavior — whether it’s an internal fault, an external hacker attack, or a self-correcting mechanism of autonomous control. Its safety logic is based on measurable parameters: power, motion status, and environmental conditions. If an abnormal mode is151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00accompanied by exceeding safety thresholds, near-accidents, or suspicious combinations of motion and environmental conditions, the management unit can automatically raise the safety thresholds or shut down smart devices.

[0082] In some embodiments, when significant anomalous behavior is detected, the management unit can log events related to time, operating environment, and sensor data for subsequent analysis. Therefore, the power management system provides a layer of protection that does not rely on the control system of the smart device for any software notifications regarding internal program changes.IV. Additional Examples

[0083] Several aspects of the present technology are set forth in the following examples:1. A power management system for a smart device, comprising:a power source electrically coupled to the smart device;a monitoring unit configured to collect power source data of the power source and behavioral data of at least one actuator of the smart device;a management unit that receives the power source data and the behavioral data from the monitoring unit and is communicatively isolated from external communication networks and the smart device; anda power controller electrically coupled to the power source and the smart device that can terminate an electrical connection between the power source and the smart device, wherein the power controller is operably coupled to the management unit, and wherein the management unit sends instructions to the power controller to terminate the electrical connection between the power source and the smart device based on abnormalities in the power source data or the behavioral data.2. The power management system of example 1 wherein the power source data includes at least one of:current data,voltage data, orpower data.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO003. The power management system of either example 1 or example 2 wherein the behavioral data of the at least one actuator includes at least one oftorque data,motor current data,rotational speed data,joint angle data,joint speed data, orjoint acceleration data.4. The power management system of any of examples 1-3 wherein the monitoring unit includes a sensor interface for directly receiving signals from a current sensor, a torque sensor, an encoder, or a joint angle sensor installed on a motor or the at least one actuator, and wherein the sensor interface is communicatively isolated from the smart device.5. The power management system of any of examples 1-4 further comprising a visual sensor positioned to monitor a workspace surrounding the smart device.6. The power management system of example 5 wherein the visual sensor is a camera electrically connected to the management unit, and wherein image data collected by the camera cannot be accessed by the smart device.7. The power management system of example 5 wherein the visual sensor monitors external objects in the workspace, and wherein the management unit includes an image analysis module configured to analyze the external objects.8. The power management system of example 7 wherein the image analysis module classifies the external objects into infants, children, or adults, and wherein the image analysis module sets different safety thresholds for each type of external object.9. The power management system of any of examples 1-8 wherein the management unit is configured to compare real-time behavioral data of the at least one actuator with an actuator behavior configuration file stored in a safety model, and determine that the smart device151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00is in an abnormal or unsafe state when a deviation between the real-time behavioral data and the actuator behavior configuration file exceeds a preset threshold.10. The power management system of example 9 wherein the management unit is configured to determine the smart device is in the abnormal or unsafe state when a sudden torque peak, oscillating motion exceeding a set threshold, prohibited directional motion, or repeated impact characteristic is detected, and instruct the power controller to reduce a speed of the smart device, reduce a torque output of the smart device, or terminate the electrical connection between the power source and the smart device.11. The power management system of any of examples 1-10 wherein the smart device includes at least one of a humanoid robot, an autonomous vehicle, unmanned aerial vehicles, or industrial robotic arms.12. The power management system of any of examples 1-11 wherein the power source, the power controller, and the management unit are housed in a power control box, and wherein an outer surface of the power control box includes a visible safety7mark that can be identified by at least one visual sensor.13. The power management system of example 12 wherein, when the at least one actuator is detected within a preset area of an image captured by the at least one visual sensor, the management unit can determine whether the at least one actuator of the smart device is approaching or contacting the power control box, and wherein the management unit is configured to determine a tampering condition is present when:the visible safety' mark is continuously detected within a distance less than a threshold distance from the at least one actuator by the at least one visual sensor for a duration exceeding a preset time:the visible safety mark and the at least one actuator are detected simultaneously by the at least one visual sensor; ora sequence of actions including loosening, cutting, or disassembling the power controller is detected.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0014. The power management system of example 13 wherein the management unit is configured to instruct the power controller to terminate the electrical connection between the power source and the smart device when tampering is detected.15. The power management system of any of examples 1-14 wherein the management unit includes a machine learning module configured to (i) store logs of the power source data, the behavioral data, and image data collected during daily operation of the power management system, (ii) train and update a safety model based on the logs, and (iii) automatically adjust at least one safety threshold for determining abnormal or unsafe states of the smart device.16. The power management system of example 15 wherein the machine learning module is further configured to process security events and near misses as training data, and optimize the safety model to reduce false positive rates, while maintaining or improving detection sensitivity to real dangerous situations of the management unit.17. The power management system of any of examples 1-16 wherein the management unit is configured to set an upper limit constraint on a safety threshold to ensure that the safety threshold does not exceed a preset safety hardware limit.18. The power management system of example 17 wherein the management unit is configured to instruct the power controller to switch to a default safe state, which includes at least reducing power to the smart device, when the management unit detects an internal fault, a missing sensor input, or a fault with a machine learning module.19. The power management system of example 17 wherein the management unit is configured to store logs of the power source data, the behavioral data, and image data before and after a security event, and record violation thresholds or rule information when a power management system shutdown or the security event occurs.20. The power management system of example 19 wherein the management unit includes a emergency stop input terminal, and wherein the management unit is configured to: immediately instruct the power controller to terminate the electrical connection between the151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00power source and the smart device and simultaneously record the security event and the power source data, the behavioral data, and the image data when the emergency stop input terminal is activated.21. The power management system of any of examples 1-20 wherein the power source, the power controller, and the management unit are housed in a power control box, and wherein the power control box includes an external touch panel mounted on the power control box and covered by a mechanical cover or door.22. The power management system of example 21 wherein the external touch panel is only operable when the mechanical cover is manually opened, and wherein the external touch panel is configured to allow an authorized operator to temporarily adjust one or more safety thresholds associated with a selected task profile.23. The power management system of example 22 wherein the one or more safety thresholds includes at least one of:maximum permissible torque,maximum permissible speed,maximum permissible acceleration,maximum permissible power,maximum permissible current, andmaximum pennissible contact force.24. The power management system of example 23 wherein the one or more safety thresholds are based on a factory training procedure executed under controlled normal operating conditions on the smart device or a reference device, and wherein the management unit is configured to automatically reset the one or more safety thresholds at a predetermined time after the authorized operator temporarily adjusts the one or more safety thresholds.25. The power management system of example 23 wherein each task profile is associated with at least one constraint selected from at least one of:maximum allowable duration,maximum number of running cycles, and151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00requirements for physical confirmation input to the external touch panel.26. The power management system of any of examples 1-25 wherein the management unit comprises a weapon disabling module configured to (i) analyze image data from at least one visual sensor to detect and classify objects within a workspace of the smart device, (ii) identify dangerous objects such as knives, guns, sticks, and baseball bats.27. The power management system of example 26 wherein the weapon disabling module is further configured to identify when an end effector or gripper of the smart device is grasping, about to grasp, a dangerous object, or the smart device moves into a position indicating that the dangerous object can be used as a weapon.28. The power management system of example 26 wherein the power source, the power controller, and the management unit are housed in a power control box with an external touch panel configured to allow an authorized operator to temporarily activate a manual weapon use authorization function, wherein the manual weapon use authorization function enables operation permissions for one or more specified weapon ty pes, and set special security' thresholds during authorized weapon operation.29. The power management system of example 28 wherein the special security thresholds are associated with at least one parameter:a maximum allowed duration,a maximum allowed number of operations or cycles,input via the external touch panel, ora specific task description.30. The power management system of example 29 wherein the management unit is configured to enforce a set of dedicated weapon operation safety thresholds that are not lower than factory security7thresholds configured for normal operation, even if a manual weapon use permit is activated.31. The power management system of example 30 wherein the factory security thresholds and associated configuration files contain a factory default state in which the151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00dangerous objects are prohibited and modifications made by the authorized operator via the manual weapon use authorization function can be considered as temporary overwrites.32. The power management system of example 31 wherein the management unit is configured to instruct the power controller to terminate the electrical connection between the power source and the smart device when the power management system detects that the smart device is simultaneously operating a dangerous object and attempting to grab, remove, damage, or open the power control box.33. The power management system of any of examples 1-32 wherein the management unit includes a machine learning module configured to identify indicators in the power source data, the behavioral data, or image data that continuously deviate from a learned normal pattern, and wherein the management unit is configured to switch to a conservative mode, imposing stricter limits on speed, torque, and power, or issuing a shutdown command for the smart device if necessary, when persistent deviations are detected.34. A method for safe pow er management of a smart device, the method comprising: collecting baseline power source data and baseline actuator behavioral data under normal operating conditions of the smart device:generating a safety model based on the baseline power source data and the baseline actuator behavioral data;collecting real-time power source data, real-time actuator behavioral data, and real-time image data during operation of the smart device;updating the safety model and threshold settings of the smart device using a machine learning module based on the real-time power source data, the real-time actuator behavioral data, and the real-time image data;determining whether there is an abnormal state of the smart device by comparing the real-time power source data, the real-time actuator behavioral data, and the realtime image data with the safety model and the threshold settings; and limiting or eliminating power to the smart device when the abnormal state is detected.35. The method of example 34, further comprising receiving temporary adjustments to the threshold settings by an authorized user.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0036. The method of either example 34 or example 35, further comprising: determining a distance and relative orientation between an end effector of the smart device and an external object;determining a safety limit based on a type of the external object and the distance between the end effector and the external object, wherein the safety limit is a maximum contact force, a maximum torque, a maximum speed, or a maximum power; and instructing a power controller to reduce power output to the smart device or perform an emergency shutdown operation of the smart device when the safely limit is exceeded or expected to be exceeded.37. The method of any of examples 34-36, further comprising:determining at least one unsafe zone around a workspace of the smart device; and instructing a power controller to reduce power output to the smart device or perform an emergency shutdown operation of the smart device when an external object enters the at least one unsafe zone.V. Conclusion

[0084] It will be apparent to those having skill in the art that changes may be made to the details of the above-described embodiments without departing from the underlying principles of the present disclosure. In some cases, well-known structures and functions have not been shown or described in detail to avoid unnecessarily obscuring the description of the embodiments of the present technology. Although steps of methods may be presented herein in a particular order, alternative embodiments may perform the steps in a different order. Similarly, certain aspects of the present technology disclosed in the context of particular embodiments can be combined or eliminated in other embodiments. Furthermore, while advantages associated with certain embodiments of the present technology may have been disclosed in the context of those embodiments, other embodiments can also exhibit such advantages, and not all embodiments need necessarily exhibit such advantages or other advantages disclosed herein to fall within the scope of the technology. Accordingly, the disclosure and associated technology can encompass other embodiments not expressly shown or described herein, and the invention is not limited except as by the appended claims.

[0085] Where the context permits, singular or plural terms may also include the plural or singular term, respectively. For example, throughout this disclosure, the singular terms “a,” “an,”151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00and “the'’ include plural referents unless the context clearly indicates otherwise. Moreover, unless the word “or” is expressly limited to mean only a single item exclusive from the other items in reference to a list of two or more items, then the use of “or” in such a list is to be interpreted as including (a) any single item in the list, (b) all of the items in the list, or (c) any combination of the items in the list. Furthermore, as used herein, the phrase “and / or” as in “A and / or B” refers to A alone, B alone, and both A and B. Additionally, the terms “comprising,” “including,” “having,” and “with” are used throughout to mean including at least the recited feature(s) such that any greater number of the same features and / or additional types of other features are not precluded. Moreover, as used herein, the phrases “based on,” “depends on,” “as a result of,” and “in response to” shall not be construed as a reference to a closed set of conditions. For example, a step that is described as “based on condition A” may be based on both condition A and condition B without departing from the scope of the present disclosure. In other words, as used herein, the phrase “based on” shall be construed in the same manner as the phrase “based at least in part on” or the phrase “based at least partially on.”

[0086] Reference herein to “one embodiment,” “an embodiment,” “some embodiments” or similar formulations means that a particular feature, structure, operation, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present technology. Thus, the appearances of such phrases or formulations herein are not necessarily all referring to the same embodiment. Furthermore, various particular features, structures, operations, or characteristics may be combined in any suitable manner in one or more embodiments.

[0087] Unless otherwise indicated, all numbers expressing numerical values used in the specification and claims, are to be understood as being modified in all instances by the term “about.” Accordingly, unless indicated to the contrary, the numerical parameters set forth in the specification and attached claims are approximations that may vary depending upon the desired properties sought to be obtained by the present technology. At the very least, and not as an attempt to limit the application of the doctrine of equivalents to the scope of the claims, each numerical parameter should at least be construed in light of the number of reported significant digits and by applying ordinary rounding techniques. The terms “about,” “approximately,” and “substantially” as used herein shall be interpreted to mean within ±10% of the stated value. Additionally, all ranges disclosed herein are to be understood to encompass the endpoints, and any and all subranges subsumed therein. For example, a range of “1 to 10” includes any and all subranges between (and including) the minimum value of 1 and the maximum value of 10 (e.g.,151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00any and all subranges having a minimum value of equal to or greater than 1 and a maximum value of equal to or less than 10. such as 5.5 to 10).

[0088] The disclosure set forth above is not to be interpreted as reflecting an intention that any claim or example requires more features than those expressly recited in that claim or example. Rather, as the preceding examples and the following claims reflect, inventive aspects lie in a combination of fewer than all features of any single foregoing disclosed embodiment. Thus, the preceding examples and the following claims are hereby expressly incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment. This disclosure includes all permutations of the independent claims with their dependent claims.

Claims

151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00CLAIMSI / W e claim:

1. A power management system for a smart device, comprising:a power source electrically coupled to the smart device;a monitoring unit configured to collect power source data of the power source and behavioral data of at least one actuator of the smart device;a management unit that receives the power source data and the behavioral data from the monitoring unit and is communicatively isolated from external communication networks and the smart device; anda power controller electrically coupled to the power source and the smart device that can terminate an electrical connection between the power source and the smart device, wherein the power controller is operably coupled to the management unit, and wherein the management unit sends instructions to the power controller to terminate the electrical connection between the power source and the smart device based on abnormalities in the power source data or the behavioral data.

2. The power management system of claim 1 wherein the power source data includes at least one of:current data,voltage data, orpower data.

3. The power management system of claim 1 wherein the behavioral data of the at least one actuator includes at least one of:torque data,motor current data,rotational speed data,joint angle data,joint speed data, orjoint acceleration data.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO004. The power management system of claim 1 wherein the monitoring unit includes a sensor interface for directly receiving signals from a current sensor, a torque sensor, an encoder, or a joint angle sensor installed on a motor or the at least one actuator, and wherein the sensor interface is communicatively isolated from the smart device.

5. The power management system of claim 1 further comprising a visual sensor positioned to monitor a workspace surrounding the smart device.

6. The power management system of claim 5 wherein the visual sensor is a camera electrically connected to the management unit, and wherein image data collected by the camera cannot be accessed by the smart device.

7. The power management system of claim 5 wherein the visual sensor monitors external objects in the workspace, and wherein the management unit includes an image analysis module configured to analyze the external objects.

8. The power management system of claim 7 wherein the image analysis module classifies the external objects into infants, children, or adults, and wherein the image analysis module sets different safety thresholds for each type of external object.

9. The power management system of claim 1 wherein the management unit is configured to compare real-time behavioral data of the at least one actuator with an actuator behavior configuration file stored in a safety model, and determine that the smart device is in an abnormal or unsafe state when a deviation between the real-time behavioral data and the actuator behavior configuration file exceeds a preset threshold.

10. The power management system of claim 9 wherein the management unit is configured to determine the smart device is in the abnormal or unsafe state when a sudden torque peak, oscillating motion exceeding a set threshold, prohibited directional motion, or repeated impact characteristic is detected, and instruct the power controller to reduce a speed of the smart device, reduce a torque output of the smart device, or terminate the electrical connection between the power source and the smart device.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0011. The power management system of claim 1 wherein the smart device includes at least one of a humanoid robot, an autonomous vehicle, unmanned aerial vehicles, or industrial robotic arms.

12. The power management system of claim 1 wherein the power source, the power controller, and the management unit are housed in a power control box. and wherein an outer surface of the power control box includes a visible safety mark that can be identified by at least one visual sensor.

13. The power management system of claim 12 wherein, when the at least one actuator is detected within a preset area of an image captured by the at least one visual sensor, the management unit can determine whether the at least one actuator of the smart device is approaching or contacting the power control box, and wherein the management unit is configured to determine a tampering condition is present when:the visible safety mark is continuously detected within a distance less than a threshold distance from the at least one actuator by the at least one visual sensor for a duration exceeding a preset time;the visible safety mark and the at least one actuator are detected simultaneously by the at least one visual sensor; ora sequence of actions including loosening, cutting, or disassembling the power controller is detected.

14. The power management system of claim 13 wherein the management unit is configured to instruct the power controller to terminate the electrical connection between the power source and the smart device when tampering is detected.

15. The power management system of claim 1 wherein the management unit includes a machine learning module configured to (i) store logs of the power source data, the behavioral data, and image data collected during daily operation of the power management system, (ii) train and update a safety model based on the logs, and (iii) automatically adjust at least one safety7threshold for determining abnormal or unsafe states of the smart device.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0016. The power management system of claim 15 wherein the machine learning module is further configured to process security events and near misses as training data, and optimize the safety model to reduce false positive rates, while maintaining or improving detection sensitivity to real dangerous situations of the management unit.

17. The power management system of claim 1 wherein the management unit is configured to set an upper limit constraint on a safety threshold to ensure that the safety threshold does not exceed a preset safely hardware limit.

18. The power management system of claim 17 wherein the management unit is configured to instruct the power controller to switch to a default safe state, which includes at least reducing power to the smart device, when the management unit detects an internal fault, a missing sensor input, or a fault with a machine learning module.

19. The power management system of claim 17 wherein the management unit is configured to store logs of the power source data, the behavioral data, and image data before and after a security event, and record violation thresholds or rule information when a power management system shutdown or the security event occurs.

20. The power management system of claim 19 wherein the management unit includes a emergency stop input terminal, and wherein the management unit is configured to: immediately instruct the power controller to terminate the electrical connection between the power source and the smart device and simultaneously record the security event and the power source data, the behavioral data, and the image data when the emergency stop input terminal is activated.

21. The power management system of claim 1 wherein the power source, the power controller, and the management unit are housed in a power control box, and wherein the power control box includes an external touch panel mounted on the power control box and covered by a mechanical cover or door.

22. The power management system of claim 21 wherein the external touch panel is only operable when the mechanical cover is manually opened, and wherein the external touch151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00panel is configured to allow an authorized operator to temporarily adjust one or more safety thresholds associated with a selected task profile.

23. The power management system of claim 22 wherein the one or more safety thresholds includes at least one of:maximum permissible torque,maximum permissible speed,maximum permissible acceleration,maximum permissible power,maximum permissible current, andmaximum permissible contact force.

24. The power management system of claim 23 wherein the one or more safety thresholds are based on a factory training procedure executed under controlled normal operating conditions on the smart device or a reference device, and wherein the management unit is configured to automatically reset the one or more safety thresholds at a predetermined time after the authorized operator temporarily adjusts the one or more safety thresholds.

25. The power management system of claim 23 wherein each task profile is associated with at least one constraint selected from at least one of:maximum allowable duration,maximum number of running cycles, andrequirements for physical confirmation input to the external touch panel.

26. The power management system of claim 1 wherein the management unit comprises a weapon disabling module configured to (i) analyze image data from at least one visual sensor to detect and classify objects within a workspace of the smart device, (ii) identify dangerous objects such as knives, guns, sticks, and baseball bats.

27. The power management system of claim 26 wherein the weapon disabling module is further configured to identify when an end effector or gripper of the smart device is grasping, about to grasp, a dangerous object, or the smart device moves into a position indicating that the dangerous object can be used as a weapon.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0028. The power management system of claim 26 wherein the power source, the power controller, and the management unit are housed in a power control box with an external touch panel configured to allow an authorized operator to temporarily activate a manual weapon use authorization function, wherein the manual weapon use authorization function enables operation permissions for one or more specified weapon ty pes, and set special safety thresholds during authorized weapon operation.

29. The power management system of claim 28 wherein the special safety thresholds are associated with at least one parameter:a maximum allowed duration,a maximum allowed number of operations or cycles,input via the external touch panel, ora specific task description.

30. The power management system of claim 29 wherein the management unit is configured to enforce a set of dedicated weapon operation safety thresholds that are not lower than factory safety thresholds configured for normal operation, even if a manual weapon use permit is activated.

31. The power management system of claim 30 wherein the factory safety thresholds and associated configuration files contain a factory default state in which the dangerous objects are prohibited and modifications made by the authorized operator via the manual weapon use authorization function can be considered as temporary’ overwrites.

32. The power management system of claim 31 wherein the management unit is configured to instruct the power controller to terminate the electrical connection between the power source and the smart device when the power management system detects that the smart device is simultaneously operating a dangerous object and attempting to grab, remove, damage, or open the power control box.

33. The power management system of claim 1 wherein the management unit includes a machine learning module configured to identity' indicators in the power source data, the behavioral data, or image data that continuously deviate from a learned normal pattern, and151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO00wherein the management unit is configured to switch to a conservative mode, imposing stricter limits on speed, torque, and power, or issuing a shutdown command for the smart device if necessary, when persistent deviations are detected.

34. A method for safe power management of a smart device, the method comprising: collecting baseline power source data and baseline actuator behavioral data under normal operating conditions of the smart device;generating a safety model based on the baseline power source data and the baseline actuator behavioral data;collecting real-time power source data, real-time actuator behavioral data, and real-time image data during operation of the smart device;updating the safety7model and threshold settings of the smart device using a machine learning module based on the real-time power source data, the real-time actuator behavioral data, and the real-time image data;determining whether there is an abnormal state of the smart device by comparing the real-time power source data, the real-time actuator behavioral data, and the realtime image data with the safety' model and the threshold settings; and limiting or eliminating power to the smart device when the abnormal state is detected.

35. The method of claim 34, further comprising receiving temporary adjustments to the threshold settings by an authorized user.

36. The method of claim 34, further comprising:determining a distance and relative orientation between an end effector of the smart device and an external object;determining a safety limit based on a type of the external object and the distance between the end effector and the external object, wherein the safety limit is a maximum contact force, a maximum torque, a maximum speed, or a maximum power; and instructing a power controller to reduce power output to the smart device or perform an emergency shutdown operation of the smart device when the safety limit is exceeded or expected to be exceeded.151234.8018.WOOO\184921248.2 Atorney Docket No. 151234.8018.WO0037. The method of claim 34, further comprising:determining at least one unsafe zone around a workspace of the smart device: and instructing a power controller to reduce power output to the smart device or perform an emergency shutdown operation of the smart device when an external object enters the at least one unsafe zone.

Images