A multi-user co-renting adaptive offline autonomous care system and method

Abstract

The application discloses a multi-user co-renting adaptive offline autonomous care system and method, and belongs to the field of intelligent old-age care robots. The system adopts a three-layer hardware architecture, and reasoning control logic local firmware runs offline, and is independent of cloud dependence. Local NPU offline prediction of falls and emergency risks and hardware active intervention; the mechanical arm is equipped with five levels of hardware redundancy protection, and fault and power failure realize physical safety bottom line. According to the self-care level of the user, a differentiated hardware acquisition strategy is set, hardware partition isolation storage, and multi-layer overwrite destruction of data when the user leaves; risk grading disposal, life risk silent directional first aid, disaster risk automatic alarm evacuation; nursing data hardware hash generation tamper-proof storage, supporting settlement and accident evidence. The application solves the privacy contradiction, safety hidden danger, rescue disturbance, network failure and settlement dispute pain points of co-renting old-age care, and realizes easy production of hardware standardization, and is suitable for co-renting, community clustering, institutional cohabitation and shared old-age care scenes.

Description

Technical Field

[0001] This invention belongs to the field of intelligent elderly care robot technology, specifically involving a local offline intelligent care system for shared nursing hardware in multi-user co-living elderly care scenarios. It is applicable to residential co-living elderly care, community-based group elderly care, and multi-person co-living scenarios in elderly care institutions, and realizes offline risk prediction, multi-level mechanical safety protection of hardware, hierarchical privacy hardware management, offline hierarchical emergency rescue, hardware isolation and destruction of user data, hardware trusted evidence storage of nursing behavior, and full-function autonomous operation when the network is disconnected. Background Technology

[0002] The existing shared smart care equipment for multiple people has many technical shortcomings:

[0003] 1. Hazard monitoring only provides passive alarms after the fact, without local proactive risk prediction based on vital signs, gait, and behavioral data. Falls and acute cardiovascular and cerebrovascular emergencies cannot be prevented through hardware intervention in advance, resulting in weak accident prevention capabilities.

[0004] 2. Shared nursing robotic arms rely solely on software algorithms for protection and limit functions. In the event of software crashes, sensor damage, or power failure, there is no hardware safety net, which can easily lead to pinching or crushing injuries.

[0005] 3. All users are subject to the same monitoring and data collection strategy, and there is no hierarchical hardware control based on their self-management capabilities, which leads to excessive collection of privacy or lack of supervision, causing disputes among roommates;

[0006] 4. In case of critical emergencies at night, the use of whole-house sound and light alarms is very likely to disturb fellow residents. Rescue instructions rely on cloud server push, and the rescue function will be directly disabled after the network is disconnected.

[0007] 5. Sharing storage areas with multiple users' data can easily lead to information leaks. Even if the data is only deleted logically by the software upon termination of the lease, the underlying storage residue can still be recovered, posing legal and privacy risks.

[0008] 6. Nursing procedures lack tamper-proof hardware records, and there is a lack of valid documentation for cost settlement and accident accountability.

[0009] 7. The overall architecture operates centrally in the cloud. After external network fluctuations or network outages, the core hardware functions for monitoring, control, and rescue cease to function.

[0010] Currently, among existing technologies and publicly available patents, there is no complete technical solution that is suitable for multi-user shared nursing hardware scenarios, integrating offline hardware risk warning, five-level progressive hardware redundancy protection, hardware hierarchical privacy management, silent offline targeted rescue, hardware physical isolation storage, hardware overwrite data destruction, hardware hash chain evidence storage, and local offline closed-loop autonomy. Summary of the Invention

[0011] Purpose of the invention

[0012] To address the industry pain points of existing shared nursing care equipment, such as weak security, privacy conflicts, disturbances during nighttime rescues, data leaks, high dependence on external networks, and disputes arising from nursing settlements, this invention provides a multi-user shared adaptive offline autonomous care system and method. This invention relies on local edge hardware to implement embedded offline AI inference, proactively predicting falls and acute health risks and intervening with hardware. It sets up five independent, progressive hardware protection levels covering everything from routine faults to extreme power outages. Hardware-level differentiated monitoring permissions are implemented based on the user's disability level. Silent offline targeted rescue is used in life-threatening situations, and hardware alarms are used for evacuation in disaster situations. The hardware bus physically isolates user storage partitions, and multi-layer hardware overwriting during lease termination achieves physical data destruction. The entire nursing process uses hardware to generate hash chain-style evidence data. All control and inference logic is embedded in local firmware, enabling self-sustaining operation even without an external network, thus solving the existing technical deficiencies in shared elderly care scenarios.

[0013] Technical solution

[0014] System Solution

[0015] The system adopts a three-layer physical hardware architecture consisting of a local edge control layer, a multimodal perception layer, and an execution terminal layer. The control program and AI inference program are embedded in the local firmware, enabling independent offline operation.

[0016] The local edge control layer is equipped with an NPU inference chip, a hardware isolated encrypted storage unit, and a hardware hash operation coprocessing unit, and has four built-in functional modules. It analyzes the data collected by the perception layer through the local NPU to calculate the risk level offline and dynamically links the sensor sampling frequency and the task scheduling priority of the nursing robotic arm.

[0017] The terminal is configured with a shared nursing mechanical actuator and a five-level independent hardware safety protection mechanism. Different operating conditions are triggered independently at each level, with physical redundancy in the hardware design. A single-level failure does not affect the operation of the other protection levels. The five-level mechanism consists of a distance reduction mechanism, a compliant force control unit, an intentionally compliant impedance module, a mechanical overload relief structure, and a power failure braking and locking mechanism, which respectively address five types of operating conditions: proximity collision, contact pressure, active force application, software failure, and complete machine power failure.

[0018] The multimodal perception layer integrates a vital signs acquisition unit, a UWB indoor positioning unit, an image acquisition module with hardware on / off locking, an environmental sensor group, an offline Bluetooth LAN module, and a local voice interaction unit; image capture permissions are controlled by the on / off state of the hardware power supply path, and the hardware power supply is only turned on after receiving local voice authorization.

[0019] The system configures differentiated hardware sensing and acquisition strategies based on different user levels: fully self-reliant, semi-self-reliant, and disabled. The hardware bus is divided into physically isolated storage partitions. When a user passes through someone else's private area, the hardware temporarily disconnects the power supply to that area and automatically restores it after the user leaves. The system builds and dynamically updates the user's digital twin model offline. Hardware resources are isolated and scheduled during multi-task concurrency. After a user terminates the lease, the dedicated storage partition is physically destroyed by performing at least three rounds of binary random overwriting.

[0020] The hardware logic distinguishes between two handling procedures for life-threatening emergencies and disaster emergencies: In life-threatening emergencies, all hardware in the house operates silently, offline Bluetooth pushes rescue information, automatically dials emergency services, issues a one-time temporary door lock password and alerts family members; in choking emergencies, only local lighting and voice emergency guidance are activated, and nearby personnel are awakened at low frequency to assist in rescue; in fire and gas leak disaster emergencies, a full-area sound and light alarm is activated, the electric door lock is unlocked, and mechanical mechanisms assist disabled users in evacuation; if the rescue contact does not confirm the response within the time limit, the alarm is automatically switched to the backup contact.

[0021] The entire nursing process involves hardware-based data collection and quality control. After comparison and evaluation with standard indicators, results are generated. The raw data is sent to a hardware hash coprocessor to generate 256-bit chained evidence data. Public and private nursing data are stored independently in hardware partitions. Multiple nursing tasks are dynamically scheduled and shared by hardware according to four priority levels: life-threatening tasks, high-risk warning tasks, user disability level, and routine nursing tasks.

[0022] Hazard identification relies on cross-verification of signals from multiple sensors: fire detection determines abnormalities in smoke and thermal imaging simultaneously; choking detection determines two or more signals lasting for 8 seconds; cardiac arrest detection determines no output of heartbeat and respiratory signals for 10 seconds, reducing the probability of false triggering.

[0023] The system assigns dedicated hardware data collection metrics to tasks such as turning over, moving, feeding, and cleaning. Private cleaning tasks have set single-time hardware permissions, and tasks are automatically postponed during sleep and bathing states. Data partitioning supports cost accounting. The system integrates an electronic fence anti-wandering module and an offline consumables prediction unit, with tiered control over out-of-bounds access and one-way data upload for consumables when insufficient. A full-process operation log generates tamper-proof, chain-like hardware-based data storage, with an anti-erasure locking structure for the storage hardware.

[0024] Methods and solutions

[0025] Steps S1 to S8 are all executed offline using hardware, completing the entire process of hardware initialization partition modeling, offline risk reasoning and tiered intervention, hardware fusion judgment of risk tiered handling, hardware quality control hash storage, digital twin dynamic update, hardware data destruction upon lease termination, self-sustaining operation without network access, and multi-task hardware priority scheduling.

[0026] Beneficial effects

[0027] 1. Three-layer pure local hardware architecture, computing power control and storage are localized and solidified, core functions run uninterrupted even when the network is offline, adaptable to elderly care scenarios in old communities with unstable networks;

[0028] 2. Five-level physical redundancy hardware protection, with hardware backup for software, sensors, and extreme power outage conditions, completely eliminates safety hazards of shared robotic arms;

[0029] 3. Hardware-level privacy isolation + partitioned physical storage + temporary power outage shielding + at least 3 rounds of hardware overwrite and destruction, eliminating privacy disputes and legal risks in co-living from the hardware level;

[0030] 4. Tiered silent hardware rescue, zero noise and light interference throughout the house in life-threatening situations, targeted emergency rescue and issuance of temporary door lock permissions, full evacuation in disaster situations, balancing rescue efficiency and rest for co-residents;

[0031] 5. Hardware-based 256-bit chained immutable evidence storage provides judicial-grade credentials for fee settlement and accident liability determination;

[0032] 6. Hardware resources are dynamically and adaptively scheduled, and the monitoring intensity and nursing priority are automatically matched according to the risk and disability level, enabling intelligent and differentiated care;

[0033] 7. All hardware is produced using standardized mass-produced hardware available on the market. It can be deployed as an integrated whole machine or as a distributed gateway in a modular manner. Mass production costs are controllable and the commercialization threshold is low. Attached Figure Description

[0034] Figure 1 Schematic diagram of the overall hardware architecture of the three-layer system;

[0035] Figure 2 Five-level progressive security protection hardware stage triggering logic block diagram;

[0036] Figure 3 User disability level and hardware privacy policy matching matrix diagram;

[0037] Figure 4 Flowchart of hardware-based physical isolation storage partitioning and physical erasure of leased data;

[0038] Figure 5 Block diagram of hardware scheduling priority and response strategy for hazard classification;

[0039] Figure 6 Flowchart for a silent, offline, and precise emergency call system for life-threatening situations throughout the house;

[0040] Figure 7 Flowchart of hardware control for multi-sensor fire hazard assessment and automatic evacuation;

[0041] Figure 8 General framework diagram for human-machine collaborative nursing hardware quality monitoring;

[0042] Figure 9 Schematic diagram of dedicated hardware monitoring indicators for various nursing tasks;

[0043] Figure 10 Flowchart for graded silent emergency resuscitation and directional awakening for nocturnal choking and cardiac arrest;

[0044] Figure 11 Flowchart of electronic fence anti-wandering control and offline predictive replenishment of consumables;

[0045] Figure 12 Flowchart of offline update and risk adaptive adjustment of user dynamic digital twin model;

[0046] Figure 13 Schematic diagram of local offline autonomous hardware architecture and one-way encrypted communication;

[0047] Figure 14 Flowchart for proactive offline prediction and intervention of multiple user falls and health risks.

[0048] Brief description of the attached diagram: Figure 1 The three layers of hardware modules and the flow of hardware signals are displayed vertically in a hierarchical manner. Figure 2 The horizontal display shows the trigger conditions and independent redundancy relationships of the five levels of hardware protection. Figure 3 A two-dimensional matrix displays the hardware on / off policies for different users and regions; Figures 4-14 All diagrams are flowcharts of hardware signal flow and hardware action execution. Arrows represent hardware control commands and data transmission directions. The technical content of the diagrams and texts completely matches the descriptions in the instruction manual. Detailed Implementation

[0049] Example 1: Five-level progressive hardware security protection extreme working condition test scenario

[0050] When a shared nursing robotic arm approaches a severely disabled user to perform a lifting action, a pre-contact deceleration mechanism is triggered at a preset distance from the user to slow down the arm. After mechanical contact with the user, a secondary compliant force control unit limits the force threshold. When the user twists their body in the opposite direction, a tertiary impedance module dynamically adjusts the torque. In case of software or sensor failure and torque overload, a tertiary mechanical structure passively releases pressure. In case of an unexpected power failure, a tertiary braking mechanism locks and slowly lowers the robotic arm. If any one of the protection power supplies is disconnected, the remaining levels can continue to operate normally independently.

[0051] Example 2: Three-user tiered privacy hardware management co-living scenario

[0052] The shared living space is configured for fully self-reliant, semi-self-reliant, and disabled users. The main control hardware is allocated physically isolated storage partitions, and cross-partition read / write is prohibited. The camera hardware in the bedroom and bathroom of fully self-reliant users is powered off when the power is off, and only vital sign monitoring is retained. The infrared monitoring in the bathroom of semi-self-reliant users is turned on, and the camera hardware in the bedroom is physically powered off. The disabled users retain only the non-visible light sensing hardware, and the high-definition visible light shooting path is permanently cut off. When a user walks through the private area of ​​another person, the power supply to that area is temporarily disconnected, and it is automatically restored after the user leaves.

[0053] Example 3: Early morning silent offline cardiac arrest emergency scenario

[0054] If the heart rate and respiratory sensors show no signal output for 10 consecutive seconds, the hardware fusion determines that cardiac arrest has occurred; the system shuts down all sound and light circuits in the house and operates silently, automatically dials the emergency number to report the location of the emergency; offline Bluetooth pushes alarms, location information, emergency guidance, and a one-time time-limited door lock password to the rescue terminal, and simultaneously alerts family members; the rescuer confirms receipt at the terminal, completing the offline precise rescue.

[0055] Example 4: Targeted local rescue for choking at night

[0056] If two types of physiological sensors emit abnormal signals for 8 seconds, the hardware determines that there is a choking hazard; the entire room remains silent and an emergency call is automatically made; an alarm is pushed to the caregiver's terminal, and only local low-brightness lighting and targeted emergency voice are turned on to wake up the nearest self-care user to assist in the rescue, while the hardware in other areas remains off; the operation data is stored in partitions and a hardware hash is generated for evidence storage.

[0057] Example 5: Dual-sensor fire detection for automatic evacuation scenarios

[0058] The smoke sensor and infrared thermal imaging sensor hardware malfunctioned, and the hardware determined that there was a fire hazard. The whole house sound and light alarm was activated, the door lock hardware was unlocked, the mechanical mechanism assisted disabled users to get up and evacuate, and self-care personnel were given voice guidance to evacuate. After the evacuation was completed, the system automatically dialed the fire alarm to report the information.

[0059] Example 6: Hardware quality control and evidence storage scenario for turning over in nursing care

[0060] The main controller issues the turning task, the hardware collects pressure distribution and body position angle data, and uploads the completion receipt upon completion of the task; the raw sensor data, evaluation results and receipt data are sent to the hardware hash coprocessor to generate chained evidence for service settlement and liability determination.

[0061] Example 7: Hardware Licensing Scenario for Single-Service Personal Cleaning

[0062] Users issue bedroom cleaning commands via voice, and the system generates one-time limited hardware access permissions. Once the task ends, the hardware lock for the permissions expires. The system automatically postpones cleaning tasks based on the user's sleep or bathing status. Public and private cleaning records are stored separately to calculate shared care costs.

[0063] Example 8: Anti-lost and offline consumable supply scenario

[0064] When a high-risk user touches the boundary of the electronic fence, the system will sequentially execute voice warning, lock the door, and issue an offline family member alarm. The system predicts consumption offline based on nursing frequency and consumable inventory. If the inventory is lower than the threshold, it will generate a replenishment data packet locally and upload it to the platform in a one-way encrypted manner after connecting to the network.

[0065] Example 9: Adaptive Scheduling Scenario for Dynamic Update of Digital Twin

[0066] When a semi-independent user's physiological and gait data continues to decline, the NPU offline inference determines that the user's physical condition is deteriorating and pushes a suggestion to change the disability level; after confirmation, the hardware automatically switches the privacy policy and increases the task priority; and an emergency warning is immediately triggered when a sudden physiological abnormality occurs.

[0067] Example 10: Hardware Data Destruction Scenario for User Termination of Service

[0068] Upon receiving the termination instruction, the hardware performs three rounds of binary random overwrite and erase on the dedicated storage partition, rendering the underlying data unrecoverable and revoking all temporary hardware permissions; the system re-optimizes the hardware task scheduling logic based on the remaining user information; and a brand-new physically isolated storage partition is allocated to new users.

[0069] Example 11: Multi-user offline risk prediction and tiered intervention scenario

[0070] The hardware continuously collects gait and vital sign data from multiple users for offline risk assessment; if the risk increases, the robotic arm is on standby and the sensor sampling frequency is increased; if the vital signs are stable, the sampling frequency is reduced to optimize privacy protection; all logs are stored in partitions and evidence is stored in hardware.

[0071] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is defined by the claims. Equivalent substitutions and modifications all fall within the scope of protection of the present invention.

[0072] Industrial applicability

[0073] The NPU inference chip, hardware hash coprocessor, physically isolated encrypted storage, nursing robotic arm, millimeter-wave radar, UWB positioning, offline Bluetooth module, and various sensors used in this invention are all standardized, mature, and commercially available components, with no customized or difficult-to-process parts. The system can be manufactured as an integrated care robot product or deployed using a distributed architecture of edge gateway + distributed sensing hardware. In the event of a network outage, all functions, including risk prediction, mechanical safety protection, privacy hardware management, emergency rescue, and quality control and evidence storage, operate stably. It is suitable for various business models such as home-based co-living elderly care, community-based group elderly care, and institutional co-living elderly care. When connected to the network, data is uploaded unidirectionally with encrypted encryption for external verification and settlement. The technology is easy to implement, highly feasible for mass production, and has significant industrial applicability.

Claims

1. A multi-user co-living adaptive offline autonomous care system, characterized in that, It includes a three-layer physical hardware architecture consisting of a local edge control layer, a multimodal perception layer, and an execution terminal layer; the system control logic and AI inference logic are embedded in the local firmware, enabling offline closed-loop autonomous operation without the need for external cloud servers; The local edge control layer is equipped with an NPU inference chip, a hardware isolated encrypted storage unit, and a hardware hash operation coprocessing unit. The firmware integrates a multi-user front-end risk prediction module, a multi-user adaptive management and control module, a hierarchical hazard scheduling module, and a human-machine collaborative service quality control and evidence storage module. The multi-user pre-risk prediction module outputs the risk level based on local offline AI inference of the data collected by the perception layer, and links the sensor sampling frequency with the scheduling priority of the nursing execution agency. The execution terminal layer is equipped with a shared nursing mechanical execution mechanism and a five-level progressive independent hardware safety protection mechanism. The multi-level protection physical redundancy is independently triggered, and the failure of a single level does not affect the operation of other levels. The multimodal perception layer integrates vital sign acquisition, indoor positioning, image acquisition, environmental monitoring, offline LAN communication, and local voice interaction hardware; the image acquisition module is configured with a hardware on / off locking structure, and the hardware is powered on only after local voice authorization; The multi-user adaptive management and control module implements hardware-level differentiated perception and collection strategies and physical isolation of storage partitions based on the user's disability level, and supports temporary power outage collection across private areas, offline digital twin modeling of user data, hardware resource isolation scheduling of concurrent tasks, and physical destruction of leased data. The hardware of the hierarchical emergency dispatch module distinguishes between two types of handling logic: life-threatening emergencies and disaster emergencies. It executes actions such as silent emergency rescue, directional alarm, global alarm, and automatic external rescue in a hierarchical manner. If no response is received within the time limit, it automatically switches to the backup alarm channel. The human-machine collaborative service quality control and evidence storage module collects quality control data from the hardware throughout the entire nursing operation process and generates chain hash evidence storage. Public and private nursing data are stored in hardware partitions, and multiple tasks are scheduled according to preset priorities to share mechanical resources.

2. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, The five-level progressive independent hardware safety protection mechanism is configured in sequence as follows: a first-level pre-contact distance deceleration mechanism, a second-level compliant force control hardware adjustment unit, a third-level active intention compliant impedance module, a fourth-level mechanical overload relief structure, and a fifth-level power failure fall braking and locking mechanism. The first two levels provide close-range collision protection, the third level adaptively adjusts the output torque, the fourth level passively relieves pressure in case of software failure, and the fifth level physically locks the lowering of the robotic arm after power failure. If any level of protection hardware fails, the remaining levels of hardware can operate independently and normally.

3. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, The hazard assessment adopts a multi-sensor hardware signal cross-fusion verification mechanism; fire assessment requires synchronous abnormality of smoke sensor and infrared thermal imaging signals; choking assessment requires two or more types of abnormal sensing hardware signals to last for more than 8 seconds; cardiac arrest assessment requires no output of heartbeat and respiratory monitoring signals for 10 seconds.

4. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, The hardware-isolated encrypted storage unit physically isolates each user's independent storage partition through a hardware bus, and cross-partition read and write is prohibited at the hardware level; when a user terminates the lease, the exclusive partition is overwritten with at least 3 rounds of binary random overwrite, making the stored data physically unrecoverable; differentiated image acquisition hardware on / off policies are configured for fully self-reliant, semi-self-reliant, and disabled users respectively: the power supply path of the image acquisition module is disconnected in the bedroom and bathroom areas of fully self-reliant users. For semi-independent users, only the non-visible light infrared monitoring hardware is activated in the bathroom area, while the bedroom camera module is physically locked and powered off; for disabled users, only the non-visible light sensing hardware is retained, while the high-definition visible light shooting path is permanently hardware-disconnected.

5. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that... The system is configured with a tiered nighttime hardware wake-up mechanism: In the event of a life-threatening emergency, the system hardware cuts off the sound and light drive circuits throughout the house to achieve silent rescue, and pushes emergency alarm information, location coordinates, emergency guidance, and a one-time temporary door lock password to the rescue terminal via offline Bluetooth; in the event of a choking hazard, directional local lighting provides voice guidance and low-frequency wake-up to the nearest user; in the event of a disaster hazard, a full-area sound and light alarm is triggered; if the contact does not confirm the response, an alarm is automatically pushed to the backup contact terminal.

6. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, The four types of nursing tasks—turning over, moving, feeding, and cleaning—are each matched with dedicated hardware data collection indicators. Turning over collects pressure distribution and body position angle data; moving collects robotic arm force control curves and impact counts; feeding collects body position and frequency of abnormal swallowing; and cleaning collects work coverage and work duration. Private area cleaning tasks use single-use hardware access permissions. The system detects sleep or bathing status and automatically postpones task execution. Public and private cleaning data are stored separately for cost accounting.

7. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, The system integrates a hardware electronic fence anti-wandering module and an offline consumables replenishment prediction unit; when a user crosses the fence, the system sequentially executes voice dissuasion, door lock locking, and offline alarm; when the consumables inventory is lower than the threshold, a replenishment data packet is generated locally and uploaded to an external platform only after connecting to the network with one-way encryption.

8. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that, Risk logs, incident records, nursing quality control records, privacy operation records, data erasure records, and consumable replenishment records are all sent to a hardware hash coprocessor to generate 256-bit chained tamper-proof data; the storage unit is equipped with a hardware anti-erasure locking structure, and the stored data has the effect of judicial electronic evidence.

9. The multi-user co-living adaptive offline autonomous care system according to claim 1, characterized in that... Multi-user concurrent nursing tasks are dynamically scheduled with hardware resources according to four priority levels: life-threatening tasks, high-risk warning tasks, user disability level, and routine nursing tasks.

10. A multi-user co-living adaptive offline autonomous care method, applied to the system described in any one of claims 1 to 9, characterized in that... All steps are performed offline on local hardware, including: S1. System hardware initialization: allocate physical isolation storage partitions according to the user's disability level, configure privacy collection policies, and build a dynamic digital twin model of the user offline; S2. Multimodal hardware acquires multidimensional data, NPU offline inference outputs fall and acute health risk levels, and adaptively adjusts sampling frequency and actuator priority; S3. Multi-sensor signal fusion determines the type of emergency. In life-threatening situations, a silent, directional, offline rescue process is executed and temporary door lock permissions are pushed to the target. In disaster-threatening situations, a full-area alarm is activated, door locks are unlocked, and mechanical mechanisms are used to assist in evacuation. S4. Nursing task hardware collects quality control data throughout the entire process, compares and evaluates the data, generates hardware chain hash evidence, and archives public and private nursing data in separate partitions. S5. Continuously update the digital twin model offline, update the disability level when the user's status declines, and trigger emergency warnings when physiological abnormalities occur. S6. After a user terminates the lease, the hardware overwrites and destroys the partition data and reclaims temporary permissions. The system then re-optimizes the task scheduling logic based on the remaining users. S7: Core functions operate independently when the network is off; only when connected to the network is the data packet for external settlement verification encrypted and uploaded in one direction. S8: When multiple tasks are running concurrently, shared mechanical hardware resources are dynamically scheduled based on four-level priority.

Images