Active robot charging method and system based on multi-modal fusion and non-line-of-sight guidance

Abstract

The application discloses a method and system for active robot charging based on multi-modal fusion and non-line-of-sight guidance, which actively searches for robots by constructing a reverse service paradigm of a mobile charging chassis, uses the non-line-of-sight penetration capability of a UWB positioning module to obtain the position of the robot under the shielding environment and guide the chassis to search, combines a laser radar to plan a global path to bypass obstacles, and switches to a visual fine alignment mode at a close distance to cooperate with a liftable platform and a magnetic suction interface to realize adaptive height adjustment and high-success-rate docking; the navigation blind area under the complex shielding environment is effectively overcome, the robot realizes full-scene autonomous recharging in a low-power or hidden state, and the energy utilization rate and the system automation level are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of robotics, specifically to an active robot charging method and system based on multimodal fusion and non-line-of-sight guidance. Background Technology

[0002] With the increasing popularity of applications such as home companionship, inspection, and security, quadruped robots have been widely adopted due to their superior mobility and environmental adaptability. However, as high-energy-consuming mobile devices, the battery life of quadruped robots has always been a key bottleneck restricting their long-term autonomous operation. Existing quadruped robot charging solutions mainly adopt the fixed charging station model. In this model, the charging station is pre-installed in a fixed location (such as a corner or a specific charging station), and the robot needs to actively plan its path using its own navigation system to move a long distance to the fixed location for wireless sensing or physical plug-in charging.

[0003] This traditional approach has significant technical drawbacks in practical applications: when a quadruped robot lies in complex, obstructed environments such as under a bed, sofa, or cabinet due to battery depletion, malfunction, or while performing a task, traditional vision and lidar navigation often fail. This causes the robot to be unable to identify the location of a fixed charging station or plan an effective return path. In this situation, the robot not only cannot complete charging but may also become disconnected due to a complete power outage, requiring manual intervention to retrieve and recharge it. This not only reduces the robot's energy utilization and operational efficiency but also severely impacts user experience and the system's automation level. Summary of the Invention

[0004] This invention aims to provide an active robot charging method and system based on multimodal fusion and non-line-of-sight guidance, in order to solve the problem in the prior art that robots cannot recharge autonomously in occluded environments and require manual intervention.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an active robot charging method based on multimodal fusion and non-line-of-sight guidance, comprising the following steps: S1. When the remaining battery power of the quadruped robot is lower than the preset threshold, the quadruped robot sends a charging request to the mobile charging chassis and activates the UWB tag. S2. After receiving a charging request, the mobile charging chassis wakes up the UWB positioning module and performs UWB ranging with the quadruped robot to obtain the distance and direction between them, thereby generating an initial search direction; wherein, when the quadruped robot is occluded, the UWB positioning module provides non-line-of-sight positioning capability to guide the mobile charging chassis. S3. The mobile charging chassis activates the lidar to perform SLAM mapping, localization and path planning according to the initial search direction, and navigates around obstacles to the area where the quadruped robot is located by combining the historical map. S4. When the mobile charging chassis approaches within one meter of the quadruped robot, switch to visual docking mode, use a visual camera to identify the magnetic charging interface or visual marker on the quadruped robot, and calculate the relative position and attitude deviation between the charging interface of the mobile charging chassis and the charging interface of the quadruped robot. S5: The mobile charging chassis adjusts its movement based on relative position and posture deviation, and adjusts the height of the charging interface through the lifting mechanism so that the magnetic interface enters the adsorption range to complete the docking. After handshake confirmation, charging is started.

[0006] Preferably, in S2, at least two UWB antennas are provided on the mobile charging chassis to estimate the quadruped robot's orientation by time difference of arrival or angle of arrival.

[0007] Preferably, in S5, the execution order of motion adjustment and height adjustment is as follows: first perform angle alignment, then perform lateral position alignment, then perform height alignment, and finally make the magnetic interface attach.

[0008] Preferably, in S5, after the magnetic interface makes contact, a safety handshake process is executed. The safety handshake process includes: low voltage detection, polarity detection, insulation detection, communication handshake, BMS confirmation, and pre-charging, after which the main charging circuit is closed.

[0009] Preferably, between S2 and S3, a step is further included to determine whether the quadruped robot is within the line of sight, including: If the LiDAR point cloud or visual camera can directly observe the quadruped robot, then proceed directly to the navigation state in step S3; otherwise, remain in the UWB guidance state in S2 until it enters the visible area.

[0010] Preferably, determining whether the quadruped robot is within the line of sight includes: A ray is projected from the position of the mobile charging chassis to the approximate position of the quadruped robot, and it is determined whether there are LiDAR point cloud obstacles on the ray; if there are no obstacles, it is determined to be within the line of sight; if there are obstacles, it is determined to be outside the line of sight.

[0011] Preferably, the charging request is sent via cloud relay communication or point-to-point communication. In point-to-point communication mode, the quadruped robot uses its own SLAM positioning, IMU, vision and map information to continuously send relative position, angle deviation, obstacle information or recommended path to the mobile charging chassis in real time to guide the mobile charging chassis.

[0012] Preferably, when the quadruped robot is in a prone position, the mobile charging chassis lowers the height of the charging interface via the lifting mechanism to accommodate the quadruped robot's charging interface.

[0013] Preferably, in steps S2 to S4, when a sudden change occurs in the direction of the UWB signal, it is determined that multipath interference exists, and filtering is performed in combination with historical trajectory, map constraints and signal stability to obtain a stable robot position estimate.

[0014] On the other hand, this invention proposes an active robot charging system based on multimodal fusion and non-line-of-sight guidance, comprising: The quadruped robot is equipped with a battery management system, a power monitoring module, a UWB tag, and a magnetic charging interface. The mobile charging chassis integrates: The UWB base station module is used for ranging and communication with the UWB tag of the quadruped robot, providing non-line-of-sight positioning guidance when the quadruped robot is occluded. LiDAR is used to perform SLAM mapping, autonomous localization, obstacle recognition, and path planning. A vision camera is used to identify the magnetic charging port or visual markers of a quadruped robot at close range. A height-adjustable platform is equipped with a magnetic docking mechanism that matches the magnetic charging interface, allowing for dynamic adjustment of the charging interface height. The autonomous navigation module controls the mobile charging chassis to move toward the quadruped robot and complete docking based on multimodal perception information.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a reverse service paradigm where a mobile charging chassis actively seeks out the robot. Utilizing the non-line-of-sight penetration capability of the UWB positioning module, it acquires the robot's location in occluded environments and guides the chassis search. Combined with LiDAR for global path planning to bypass obstacles, it switches to a visual precision alignment mode at close range, working with a liftable platform and magnetic interface to achieve adaptive height adjustment and high-success-rate docking. This effectively overcomes navigation blind spots in complex occluded environments, enabling the robot to autonomously recharge in all scenarios, even when low on battery or in concealed states, significantly improving energy utilization and system automation. Attached Figure Description

[0016] Figure 1 This is a flowchart of the active robot charging method based on multimodal fusion and non-line-of-sight guidance of the present invention; Figure 2 This is a block diagram of the active robot charging system based on multimodal fusion and non-line-of-sight guidance of the present invention. Figure 3This is a schematic diagram of the structure of the mobile charging chassis of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indicators will also change accordingly.

[0019] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0020] like Figure 1 As shown, this invention proposes an active robot charging method based on multimodal fusion and non-line-of-sight guidance, comprising the following steps: S1. When the remaining battery power of the quadruped robot is lower than the preset threshold, the quadruped robot sends a charging request to the mobile charging chassis and activates the UWB tag. The charging request is sent through cloud relay communication or point-to-point communication. In point-to-point communication, the quadruped robot uses its own SLAM positioning, IMU, vision and map information to continuously send relative position, angle deviation, obstacle information or recommended path to the mobile charging chassis in real time to guide the mobile charging chassis.

[0021] By utilizing cloud-based or point-to-point communication mechanisms, and especially by leveraging the quadruped robot's own SLAM, IMU, and vision-based multi-source perception information to provide real-time feedback on relative position and obstacle data, the mobile charging chassis can acquire more accurate target status and environmental awareness. This not only effectively solves the guidance lag problem caused by dynamic environmental changes in traditional one-way positioning, but also significantly improves navigation efficiency and obstacle avoidance safety in complex scenarios, ensuring rapid and reliable autonomous docking even in emergency situations with low battery.

[0022] S2. After receiving a charging request, the mobile charging chassis wakes up the UWB positioning module and performs UWB ranging with the quadruped robot to obtain the distance and direction between them, thereby generating an initial search direction. When the quadruped robot is occluded, the UWB positioning module provides non-line-of-sight positioning capability to guide the mobile charging chassis. At least two UWB antennas are set on the mobile charging chassis to estimate the quadruped robot's orientation by time difference of arrival or angle of arrival.

[0023] By utilizing multi-antenna UWB ranging and time difference of arrival / angle algorithm, the quadruped robot can still accurately obtain its orientation and distance even when it is occluded by obstacles such as furniture, effectively overcoming the problem of traditional vision or lidar failing in non-line-of-sight environments. This not only provides a reliable initial search direction for the mobile charging chassis and significantly shortens the pathfinding time in blind spots, but also significantly improves the system's positioning robustness and recharging success rate in complex occlusion scenarios.

[0024] After receiving the charging command, the mobile chassis wakes up the system and first determines whether the robot is within the line of sight (LOS): if it is within the line of sight, it directly calls the LiDAR to plan the path; if it is outside the line of sight (such as being blocked or crossing a room), it activates the UWB module and uses its centimeter-level wall-penetrating characteristics to obtain the robot's precise azimuth and distance (NLOS guidance).

[0025] The judgment method is as follows: 1. Gaze detection based on visual cameras; 1.1 Start Camera Detection: The chassis near-field vision camera is activated to detect the target robot (by tag, shape features, or charging port) and determine whether the recognition is successful. If the camera can detect the charging port or the robot body (such as AprilTag, ArUco, or depth map features) in N consecutive frames, the LOS is determined to be successful; otherwise, it is determined to be occluded, and the system enters NLOS mode.

[0026] ; Where Detection_Confidence is the confidence level of target recognition, and Threshold is a set threshold (e.g., 80%).

[0027] 2. LiDAR-based line-of-sight determination; Point cloud line visual detection: A ray is projected from the chassis position to the robot's approximate position to determine whether there are obstacle point clouds blocking the ray. The judgment logic is: if the ray is not blocked by obstacle point clouds, the LOS is valid; if it is blocked by obstacle point clouds, the LOS is invalid (i.e., NLOS). Where Ray(P_base,P_robot) is the ray from the chassis to the robot's coarse positioning point, and P_i is the LiDAR point cloud detected along the ray.

[0028] 3. Multi-sensor fusion judgment; S3: The mobile charging chassis activates the LiDAR to perform SLAM mapping, localization, and path planning based on the initial search direction, and navigates around obstacles to the area where the quadruped robot is located by combining the historical map. By integrating real-time SLAM data from LiDAR with historical map data, the mobile charging chassis can build a high-precision environmental model and plan the optimal obstacle avoidance path. This not only effectively solves the navigation interruption problem caused by dynamic obstacles and ensures accurate and safe arrival at the target area in complex and ever-changing environments, but also significantly improves the adaptability of long-distance navigation and overall operational efficiency.

[0029] In another embodiment, navigation mapping can be achieved using a depth camera.

[0030] S4. When the mobile charging chassis approaches within one meter of the quadruped robot, switch to visual docking mode, use a visual camera to identify the magnetic charging interface or visual marker on the quadruped robot, and calculate the relative position and attitude deviation between the charging interface of the mobile charging chassis and the charging interface of the quadruped robot. Preferably, in steps S2 to S4, when a sudden change occurs in the direction of the UWB signal, it is determined that multipath interference exists, and filtering is performed in combination with historical trajectory, map constraints and signal stability to obtain a stable robot position estimate.

[0031] By employing a cascaded strategy of long-range UWB coarse positioning and close-range visual precision alignment, the system utilizes a visual camera to accurately identify the magnetic interface and calculate relative attitude deviations, thus solving the problem of contact failure caused by minute positional errors in mechanical docking. Simultaneously, the introduction of multipath interference detection and fusion filtering mechanisms effectively suppresses signal jumps in complex environments, ensuring the continuity and reliability of positioning throughout the entire process from search to final high-precision docking, and significantly improving the alignment success rate of the charging interface.

[0032] S5. The mobile charging chassis adjusts its movement based on relative position and posture deviation, and adjusts the height of the charging interface through a lifting mechanism, so that the magnetic interface enters the adsorption range to complete the docking. After handshake confirmation, charging begins. The execution sequence of movement adjustment and height adjustment is as follows: first, angle alignment, then lateral position alignment, then height alignment, and finally, the magnetic interface is adsorbed.

[0033] After the magnetic interface makes contact, a safety handshake process is executed, which includes: low voltage detection, polarity detection, insulation detection, communication handshake, BMS confirmation, and pre-charging, after which the main charging circuit is closed.

[0034] Furthermore, when the quadruped robot is in a prone position, the mobile charging chassis lowers the height of the charging interface via a lifting mechanism to accommodate the quadruped robot's charging interface.

[0035] By adopting an orderly and coordinated adjustment strategy of angle, lateral, and height, combined with an adaptive lifting mechanism for the quadruped robot's posture (such as lying down), the problem of precise docking in multi-degree-of-freedom space is effectively solved, and the fault tolerance and success rate of magnetic connection are greatly improved. At the same time, through a safety handshake process that includes low voltage, polarity, insulation and BMS confirmation, electrical hazards such as short circuits and reverse connections are completely eliminated before closing the main circuit, ensuring high safety and system stability during the charging process.

[0036] Preferably, between S2 and S3, there is also a step to determine whether the quadruped robot is within the line of sight, including: if the LiDAR point cloud or the visual camera can directly observe the quadruped robot, then directly enter the navigation state of step S3; if not, then remain in the UWB guidance state of S2 until it enters the visible area.

[0037] The process of determining whether the quadruped robot is within the line of sight includes: projecting a ray from the position of the mobile charging chassis to the approximate position of the quadruped robot, and determining whether there are LiDAR point cloud obstacles on the ray; if there are no obstacles, it is determined to be within the line of sight; if there are obstacles, it is determined to be outside the line of sight.

[0038] The navigation strategy is intelligently switched through a line-of-sight detection mechanism. When the target is visible, LiDAR is used directly for efficient and accurate navigation, while when the target is obstructed, the system automatically reverts to UWB guidance mode. This dynamic adaptation effectively avoids navigation failure or blind search caused by visual / radar failure due to obstruction. It significantly shortens the pathfinding time in complex environments and greatly improves the robustness and task continuity of the system in different scenarios.

[0039] On the other hand, this invention proposes an active robot charging system with multimodal fusion and non-line-of-sight guidance, such as... Figure 2 As shown, it includes: The quadruped robot is equipped with a battery management system, a power monitoring module, a UWB tag, and a magnetic charging interface. like Figure 3 As shown, the mobile charging chassis uses a four-wheel differential drive and measures approximately 50cm × 50cm × 40cm (height adjustable after raising and lowering). The robot end integrates magnetic charging ports on the abdomen / leg sides (compatible with prone, sitting, or standing postures). UWB employs TWR (two-way ranging) + TDOA (time difference of arrival) algorithms to achieve relative positioning accuracy <10cm. It integrates: The UWB base station module is used for ranging and communication with the UWB tag of the quadruped robot, providing non-line-of-sight positioning guidance when the quadruped robot is occluded. LiDAR 1 is used to perform SLAM mapping, autonomous localization, obstacle recognition, and path planning. Visual camera 2 is used to identify the quadruped robot's magnetic charging port 4 or visual markers at close range; The liftable platform 3 is equipped with a magnetic docking mechanism that matches the magnetic charging interface, which is used to dynamically adjust the height of the charging interface. The autonomous navigation module controls the mobile charging chassis to move toward the quadruped robot and complete docking based on multimodal perception information.

[0040] Furthermore, the detailed implementation method of multimodal localization and navigation is as follows: 1. UWB long-range positioning and occlusion guidance When the quadruped robot's battery level drops below a preset threshold (e.g., 10%), the robot sends a charging request to the mobile charging chassis via WiFi, Bluetooth, 4G, or a local area network, and simultaneously activates its UWB tag. The mobile charging chassis is equipped with at least one UWB base station module, preferably with two or three UWB antennas, to acquire relative distance and direction information.

[0041] Implementation process: The quadruped robot sends a low battery charging request; the mobile charging chassis wakes up the UWB positioning module; UWB distance measurement is performed between the chassis and the robot; the distance between them is calculated using the two-way time-of-flight ranging method.

[0042] If the chassis is equipped with multiple UWB antennas, the robot's location can be estimated using the time difference of arrival or angle of arrival. Even when the robot is obstructed by sofas, chairs, corners, beds, etc., preventing direct observation by the visual camera and LiDAR, UWB can still provide approximate direction and distance. The chassis generates an initial search direction based on the UWB results and uses historical maps to navigate around obstacles.

[0043] Basic principle of ranging: UWB uses two-way time-of-flight ranging, and the calculation formula is as follows: (where d is the distance between the chassis and the robot, c is the electromagnetic wave propagation speed, and t is the round-trip time of the UWB signal). If multiple UWB receiving antennas are used, the robot's orientation angle relative to the charging chassis can be obtained based on the distance difference or angle of arrival measured by the multiple antennas.

[0044] 2. LiDAR-based mid-range SLAM mapping, localization, and path planning After obtaining the robot's approximate location using UWB, the mobile charging chassis activates the LiDAR for environmental perception, SLAM mapping, and autonomous navigation. The LiDAR is primarily responsible for obstacle recognition, map building, autonomous localization, path planning, and dynamic obstacle avoidance within the home environment.

[0045] Implementation process: The chassis activates the LiDAR to scan the surrounding environment; acquires the contour point cloud of room walls, furniture, table legs, sofas, cabinets, etc.; constructs an indoor 2D or 3D map using the SLAM algorithm; determines the approximate area where the robot is located by combining the target direction given by UWB; the chassis plans a passable path from the current position to the target area; continuously updates the map and its own pose using LiDAR during the journey; if dynamic obstacles (such as people, pets, or moving chairs) appear in the path, the path is replanned.

[0046] The implementation methods of SLAM and navigation are as follows: 2D LiDAR + GMapping / Cartographer / Hector SLAM; 3D LiDAR + LOAM / LIO-SAM.

[0047] The system can establish the following coordinate systems: world / room coordinate system W, mobile charging chassis coordinate system C, quadruped robot coordinate system R, and charging port coordinate system P. LiDAR is responsible for determining the chassis's pose in the world coordinate system. UWB provides the robot's approximate target position relative to the chassis or world coordinate system. The mobile chassis is based on the current position (PC) and the target position (PR). The environment map constructed with LiDAR is used to plan the path.

[0048] 3. Visual camera close-range recognition and precise docking When the mobile charging chassis moves close to the robot (e.g., within 1 meter), the system switches to a vision-based precision docking mode. The vision camera is primarily used to identify the magnetic charging interface, charging markers, visual tags, or structural features on the robot's end, and to calculate the relative position and attitude deviation between the chassis charging interface and the robot's charging interface. Possible identification methods include any one or more combinations of the following: Visual tag recognition: Place AprilTag, ArUco code or custom pattern near the robot's charging port, and the camera will calculate the relative pose after recognizing the tag; Charging port structure recognition: The charging port outline, groove, and magnetic terminal are identified by a depth camera, and the center point of the charging port is determined based on the shape features. Active light-assisted recognition: Infrared supplementary lights or structured light are set on the chassis to improve recognition stability in low-light environments; The robot actively displays guidance: The robot displays docking marks on the screen or indicator lights, and the chassis camera recognizes the illuminated marks to complete the alignment.

[0049] Implementation process: The chassis enters the area near the robot; the vision camera starts searching for the charging port; the robot's charging port or visual marker is identified; the center point of the charging port is calculated; the lateral deviation, longitudinal distance, and angular deviation between the chassis interface and the robot interface are calculated; the chassis makes micro-motion adjustments; the lifting mechanism adjusts the height of the charging port; the magnetic interface enters the adsorption range; charging starts after the contact handshake is successful.

[0050] Visual pose calculation is as follows: If visual labels are used, the relative pose between the camera and the label can be calculated using the PnP algorithm. Where C represents the chassis camera coordinate system, and P represents the robot charging port or tag coordinate system. After transforming the camera coordinate system to the chassis coordinate system, the deviation between the chassis charging interface and the robot charging interface can be obtained. This includes Δx (left-right deviation), Δy (front-rear distance deviation), Δz (height deviation), and Δθ (angle deviation). The chassis performs closed-loop control based on the deviations, using the following formula: ,in, For lateral correction speed, ω is the forward and backward correction velocity, h is the angular velocity, and h is the target height of the lifting platform. To control the gain.

[0051] 4. Fusion of three sensory modes Phase 1: Long-distance search phase (distance greater than 3 meters or when occlusion exists). The main sensor is UWB, and the auxiliary sensor is LiDAR; vision is off or operating at low frequency, mainly used to determine the robot's approximate direction, determine the presence of occlusion, and plan the initial path.

[0052] Phase Two: Mid-range navigation phase (approximately 1 to 3 meters away). The main sensor is LiDAR, and the auxiliary sensor is UWB; vision is in a low-frequency search state, mainly used to achieve obstacle avoidance, approach the robot's location, and determine accessible paths.

[0053] Phase 3: Close-range docking phase (within approximately 1 meter). The main sensor is a vision camera, and the auxiliary sensor is LiDAR / UWB; the actuator is a lifting mechanism + magnetic interface, mainly used to identify the charging port, adjust the posture, complete the magnetic docking, and start charging after handshake confirmation.

[0054] 5. Obstruction treatment When the robot is obstructed by furniture, preventing the chassis camera from seeing the charging port and the LiDAR from directly recognizing the robot's outline, the system enters NLOS (Non-Line-of-Sight) guidance mode. The process is as follows: UWB continuously outputs the robot's relative distance and direction.

[0055] 1. LiDAR determines obstacle boundaries based on the map; 2. The chassis moves along the passable area in the direction where the UWB distance decreases; 3. If the direction of the UWB signal changes abruptly, it indicates the presence of multipath interference; 4. Filtering is performed by combining historical trajectory, map constraints, and signal stability; 5. Once you reach the visible area, switch to visual docking mode.

[0056] The filtering method can employ Kalman filtering, extended Kalman filtering, or particle filtering to fuse UWB, LiDAR odometry, and IMU information. The formula is as follows: Where x and y represent the chassis position, θ represents the chassis orientation, and v represents the velocity. The estimated target position after fusion is: .

[0057] Charging port alignment and magnetic docking control: In the close-range stage, the chassis performs step-by-step alignment based on the deviation calculated visually, in the following order: angle alignment, lateral position alignment, height alignment, slow forward movement, magnetic assisted adsorption, electrical handshake, and charging activation.

[0058] Safety logic: After the magnetic contact makes contact, a large current is not output immediately. Instead, low voltage detection, polarity detection, insulation detection, communication handshake, BMS confirmation, pre-charging are performed in sequence. Finally, the main charging circuit is closed and the process ends.

[0059] The aforementioned quadruped robot is equipped with a battery management system (BMS) and a power monitoring module, which can monitor the following parameters in real time: remaining power (SOC), battery voltage, battery temperature, state of health (SOH), current power consumption, and remaining runtime. When the system detects that the power level is lower than a preset threshold (e.g., SOC ≤ 10%), it will trigger an automatic charging request process.

[0060] Specifically, the cloud-based communication relay method is as follows: The quadruped robot connects to the cloud server via WiFi / 4G / 5G, and the cloud server then connects to the mobile charging chassis via WiFi / 4G / 5G. When the robot detects low battery, it generates a charging request and packages the following data: {Robot_ID, SOC, current coordinates, current orientation, whether it is occluded, environment map ID, request timestamp, priority}, and then sends the data packet to the cloud. After receiving the data, the cloud performs the following processing: identity authentication (verifying the legality and authorization status of the robot and charging equipment), device scheduling (if multiple chassis exist, the nearest available chassis is selected first), and environmental information synchronization (distributing home map, historical charging points, optimal path, and other data).

[0061] The specific point-to-point communication method is as follows: the quadruped robot communicates directly with the mobile charging chassis, without relying on the cloud, making it suitable for scenarios with network outages, offline home environments, and high real-time and privacy requirements. The implementation method is as follows: Step 1: Low battery trigger. When the robot detects that its SOC is ≤10%, it enters active assistance mode.

[0062] Step 2: Broadcast a charging request. The robot periodically broadcasts a data packet containing the following information: Robot_ID, remaining battery power, current location, UWB status, posture information, and whether it is movable. Broadcast methods can include WiFi Direct, BLE broadcast, LAN UDP broadcast, UWB Beacon, ZigBee, or Thread.

[0063] Step 3: The mobile charging chassis receives the request. After receiving the broadcast, the chassis sequentially determines whether the request comes from a legitimate device, whether it is in an idle state, and whether its remaining power is sufficient to meet the charging requirements.

[0064] Step 4: Establish a point-to-point connection. This can be achieved using methods such as TCP / IP, UDP, BLE GATT, or UWB bidirectional ranging links.

[0065] Step 5: The robot guides the chassis in real time. The robot uses its own SLAM localization, IMU, vision and map data to continuously send relative position, angle deviation, obstacle information and recommended path to the chassis, such as: "Move the chassis 20cm to the left", "Move forward 50cm", "Turn 15° to the right".

[0066] In summary, by constructing a reverse service paradigm where a mobile charging chassis actively seeks out the robot, the non-line-of-sight penetration capability of the UWB positioning module is utilized to obtain the robot's location in occluded environments and guide the chassis search. Combined with LiDAR for global path planning to bypass obstacles, and switching to a visual precision alignment mode at close range, along with a liftable platform and magnetic interface, adaptive height adjustment and high-success-rate docking are achieved. This effectively overcomes navigation blind spots in complex occluded environments, enabling the robot to autonomously recharge in all scenarios, even when low on battery or in concealed states, significantly improving energy utilization and system automation.

[0067] The above description is merely an exemplary embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention specification and drawings under the technical concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. An active robot charging method based on multimodal fusion and non-line-of-sight guidance, characterized in that, Includes the following steps: S1. When the remaining battery power of the quadruped robot is lower than the preset threshold, the quadruped robot sends a charging request to the mobile charging chassis and activates the UWB tag. S2. After receiving a charging request, the mobile charging chassis wakes up the UWB positioning module and performs UWB ranging with the quadruped robot to obtain the distance and direction between them, thereby generating an initial search direction; wherein, when the quadruped robot is occluded, the UWB positioning module provides non-line-of-sight positioning capability to guide the mobile charging chassis. S3. The mobile charging chassis activates the lidar to perform SLAM mapping, localization and path planning according to the initial search direction, and navigates around obstacles to the area where the quadruped robot is located by combining the historical map. S4. When the mobile charging chassis approaches within one meter of the quadruped robot, switch to visual docking mode, use a visual camera to identify the magnetic charging interface or visual marker on the quadruped robot, and calculate the relative position and attitude deviation between the charging interface of the mobile charging chassis and the charging interface of the quadruped robot. S5: The mobile charging chassis adjusts its movement based on relative position and posture deviation, and adjusts the height of the charging interface through the lifting mechanism so that the magnetic interface enters the adsorption range to complete the docking. After handshake confirmation, charging is started.

2. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, In S2, at least two UWB antennas are installed on the mobile charging chassis to estimate the quadruped robot's orientation by time difference of arrival or angle of arrival.

3. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, In S5, the execution order of motion adjustment and height adjustment is as follows: first perform angle alignment, then perform lateral position alignment, then perform height alignment, and finally make the magnetic interface attach.

4. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, In S5, after the magnetic interface makes contact, a safety handshake process is executed. The safety handshake process includes: low voltage detection, polarity detection, insulation detection, communication handshake, BMS confirmation, and pre-charging, after which the main charging circuit is closed.

5. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, Between S2 and S3, there is also a step of determining whether the quadruped robot is within the line of sight, including: If the LiDAR point cloud or visual camera can directly observe the quadruped robot, then proceed directly to the navigation state in step S3; otherwise, remain in the UWB guidance state in S2 until it enters the visible area.

6. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 5, characterized in that, The determination of whether the quadruped robot is within the line of sight includes: A ray is projected from the position of the mobile charging chassis to the approximate position of the quadruped robot, and it is determined whether there are LiDAR point cloud obstacles on the ray; if there are no obstacles, it is determined to be within the line of sight; if there are obstacles, it is determined to be outside the line of sight.

7. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, The charging request is sent via cloud relay communication or point-to-point communication. In point-to-point communication mode, the quadruped robot uses its own SLAM positioning, IMU, vision and map information to continuously send relative position, angle deviation, obstacle information or recommended path to the mobile charging chassis in real time to guide the mobile charging chassis.

8. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, When the quadruped robot is in a prone position, the mobile charging chassis lowers the height of the charging interface through the lifting mechanism to fit the charging interface of the quadruped robot.

9. The active robot charging method based on multimodal fusion and non-line-of-sight guidance according to claim 1, characterized in that, In steps S2 to S4, when a sudden change occurs in the direction of the UWB signal, it is determined that multipath interference exists, and filtering is performed by combining historical trajectory, map constraints and signal stability to obtain a stable robot position estimate.

10. An active robot charging system based on multimodal fusion and non-line-of-sight guidance for implementing the method as described in any one of claims 1-9, characterized in that, include: The quadruped robot is equipped with a battery management system, a power monitoring module, a UWB tag, and a magnetic charging interface. The mobile charging chassis integrates: The UWB base station module is used for ranging and communication with the UWB tag of the quadruped robot, and provides non-line-of-sight positioning guidance when the quadruped robot is occluded. LiDAR is used to perform SLAM mapping, autonomous localization, obstacle recognition, and path planning. A vision camera is used to identify the magnetic charging port or visual markers of a quadruped robot at close range. A height-adjustable platform is equipped with a magnetic docking mechanism that matches the magnetic charging interface, allowing for dynamic adjustment of the charging interface height. The autonomous navigation module controls the mobile charging chassis to move toward the quadruped robot and complete docking based on multimodal perception information.

Images