A zigbee-based exoskeleton sensor networking method

Abstract

The application relates to the field of electric communication technology, in particular to a kind of exoskeleton sensor networking method based on Zigbee, steps include: master control establishes network and allocates short address for each node;All nodes are constructed to left-right symmetrical twin tree topology with master control as root;Master control periodically sends probe frame to each node, collects the received signal strength, link quality indication and end-to-end delay of each node, and simultaneously obtains the current bending angle value of important joint;Dynamic attenuation compensation factor is calculated, and signal strength is compensated and corrected, then left-right limb symmetry confidence offset is calculated, and signal strength is secondarily corrected;The comprehensive communication confidence score of each node is calculated, direct communication, cluster tree forwarding or trigger routing repair is selected according to the score;Master control dynamically adjusts the data reporting interval of the node according to phase;After double check, each node reports the collected data.The application can maintain exoskeleton stable, and low delay data communication of master control and sensor.

Description

Technical Field

[0001] This invention relates to the field of telecommunications technology, and in particular to a Zigbee-based exoskeleton sensor networking method. Background Technology

[0002] Exoskeleton robots require sensors distributed throughout the body to perceive human movement in real time. To achieve optimal control, a complete exoskeleton system typically needs nodes at multiple movable joints, such as the hip, knee, and ankle, to operate simultaneously. Sensors need to be distributed in various locations, and the wiring harnesses at joint pivot points are particularly difficult to arrange, easily leading to fatigue damage during repetitive movement. Increasing the number of sensors results in increased wiring, impacting wearability and product reliability. Using Zigbee wireless networking avoids the problems caused by wiring harnesses, enabling flexible networking of multiple sensors. While traditional Zigbee networking has sufficient node capacity, its routing mechanism and sleep scheduling do not consider the impact of the exoskeleton's unique symmetrical left and right limb structure and the periodic changes in joint angles during walking on wireless signal quality. This leads to increased data packet loss and frequent retransmissions when signal obstruction is severe, further increasing communication latency and node power consumption, failing to meet the low latency and high reliability requirements of exoskeleton closed-loop control. Summary of the Invention

[0003] This invention provides a Zigbee-based exoskeleton sensor networking method, which can effectively solve the problems in the background technology.

[0004] This invention provides a Zigbee-based exoskeleton sensor networking method, comprising: S1: The master controller scans all Zigbee channels and selects the channel with the least interference to establish the network, and assigns short addresses to each node according to the limb side and joint level depth; S2: Construct a left-right symmetrical twin-tree topology with the master node as the root of all nodes; S3: The master controller periodically sends probe frames to each node to collect the received signal strength, link quality indication and end-to-end delay of each node, and at the same time obtains the current bending angle value of important joints; S4: The main controller uses the bending angle value to calculate the dynamic attenuation compensation factor of human tissue for wireless signal and compensates and corrects the signal strength. Then, it combines the corresponding joint angle of the opposite limb to calculate the symmetrical confidence offset of the left and right limbs and performs secondary correction on the compensated signal strength. S5: The master controller uses the signal strength after secondary correction to calculate the comprehensive communication confidence score of each node, and selects direct communication, cluster tree forwarding or triggering route repair based on the score; S6: The master controller periodically broadcasts a synchronization beacon. After receiving the beacon, each node determines its current gait phase based on its own plantar pressure resultant force value and dynamically adjusts its data reporting interval according to the phase. S7: Each node performs local moving average filtering and zero drift correction on the collected sensor data, encapsulates the data frame, performs double verification, and then reports it.

[0005] Furthermore, in step S1, establishing the network specifically involves: After the main controller is powered on, it scans all Zigbee channels in the 2.4GHz band one by one and measures the average received energy value E on each channel. ch Select E ch The smallest channel is used as the working channel, and network establishment beacon frames are sent on this channel. After powering on, each node continuously scans the network beacons on the working channel. Upon receiving a beacon, it sends a network access request frame to the master controller. The network access request frame carries the node's own pre-configured attribute code, which consists of three parts: the first part identifies the left or right limb, the second part identifies the joint level to which the node belongs, and the third part identifies the node type.

[0006] Furthermore, in step S1, the allocation of the short address specifically involves: After receiving the network access request, the master controller reads the limb identifier and level identifier in the attribute code and allocates a short address from the address pool of the corresponding limb subnet: the address pool range of the left limb is 0x0000 to 0x0007, and the address pool range of the right limb is 0x0008 to 0x000F. The allocation is carried out in the order of level depth from shallow to deep, and within the same level, the allocation is carried out in the order of arrival of the network access request. After allocation, the master controller records the short address and the limb identifier and hierarchical depth binding in the attribute code in the local address mapping table, and encapsulates the short address in the network access response frame and returns it to the corresponding node. After receiving the response frame, the node saves the short address and replies with a network access confirmation frame to complete the network access. If the master controller has no available addresses in the address pool of a certain limb subnet, it will sequentially allocate the next address from the reserved extended address pool 0x0010 to 0x001F and mark the address as the extended address of that limb subnet.

[0007] Furthermore, in step S2, constructing the bisubtree topology specifically involves: The topology depth value of the master controller is marked as 0, and the limb side is marked as trunk; The sensors located on the left and right thighs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 1 for both. The sensors located on the left and right lower legs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 2 for both. The sensors located on the left heel, left toe, right heel, and right toe are labeled as left limb terminal nodes and right limb terminal nodes, respectively, with a topology depth value of 3 for both. The connection relationship between parent and child links is established step by step according to the topology depth; no direct communication link is established between the left and right limbs.

[0008] Furthermore, in step S3, the specific steps for obtaining data are as follows: The master controller sends probe frames to each node sequentially at a set period. After receiving the probe frame, each node measures the received signal strength and link quality indication of the frame and uses the measurement results as a reply frame. If the node is located at the left thigh, right thigh, left calf, or right calf, the node's joint angle value at that time is also read and filled into the reply frame. Then, all reply frames are sent back to the master control. After receiving the reply frames from each node, the master controller subtracts the sending timestamp in the probe frame from the current time to obtain the delay of that node; In step S4, the calculation of the dynamic attenuation compensation factor and the compensation correction of the signal strength are specifically as follows: Calculate the dynamic attenuation compensation factor (DAF) of the i-th node. i : DAF i =1+[sin(θ i [×π / 180)] 2 ×λ; In the formula, i takes the values ​​hip,L, hip,R, knee,L and knee,R, which represent the nodes at the left thigh, right thigh, left calf and right calf, respectively. λ is the human tissue occlusion coefficient; Calculate the equivalent signal strength of the i-th node: RSSI corr,i =RSSI raw,i +10×log 10 (DAF i ); In the formula, RSSI raw,i Let be the received signal strength of the i-th node.

[0009] Further, in step S4, the calculation of the confidence offset between the left and right limbs and the secondary correction of the compensated signal strength are specifically as follows: The main controller reads the hip joint angle θ of both sides of the body at the same moment. hip,L and θ hip,R and knee joint angle θknee,L and θ knee,R Calculate the confidence offset Δ between the left and right limbs. sym : Δ sym =(θ hip,L -θ hip,R )-(θ knee,L -θ knee,R ); Calculate the symmetric confidence correction factor (MCF): MCF=exp(-|Δ sym | / δ); In the formula, δ is the symmetry attenuation constant; Calculate the final signal strength value RSSI of the i-th node. joint,i : .

[0010] Further, in step S5, the calculation of the comprehensive communication confidence score for each node is specifically as follows: Calculate the overall communication confidence score Φ of the i-th node. i : Φ i =[1+(RSSI joint,i -R thr ) / |R thr |]×(LQI i / L thr )×exp(-D i / D thr )×Γ i ; In the formula, LQI i D is the link quality indicator for the i-th node; i R is the delay of the i-th node; thr L is the signal strength reference threshold; thr D is the baseline threshold for link quality indication; thr Γ is the delay baseline threshold; i is the historical confidence inheritance factor of the i-th node.

[0011] Furthermore, in step S5, the specific choices for direct communication, cluster-tree forwarding, or triggered route repair are as follows: When Φ i When the value is ≥0.75, the direct communication mode is selected, and the source node sends the data frame directly to the target node; When 0.45≤Φ i When the value is less than 0.75, the cluster tree routing mode is selected, and the source node forwards the data level by level along the parent-child hierarchy. When Φ iWhen the value is less than 0.45, the route repair mode is triggered. The master controller traverses all the candidate adjacent nodes of the node, calculates the Φ value of each candidate node, selects the one with the highest Φ as the new path, and generates a route repair command to send to the source node.

[0012] Furthermore, in step S6, the dynamic adjustment of the gait phase determination and reporting interval is specifically as follows: The pressure values ​​at four points—left heel, left toe, right heel, and right toe—are collected. The heel pressure value and toe pressure value on the same side of the limb are summed to obtain the resultant plantar pressure value on that side of the limb. Before broadcasting the synchronization beacon, the master controller acquires the combined plantar pressure values ​​of both sides and takes the larger one as the effective plantar pressure value F for that limb. eff ; Set the judgment threshold F thr When F eff ≥F thr When it is determined that the side of the limb is in the standing phase, when F eff <F thr The system determines if the device is in the swing phase; the reporting interval for the standing phase node is adjusted to the baseline value × 1.2, and the reporting interval for the swing phase node is adjusted to the baseline value × 0.8; the main controller adjusts the reporting interval of all nodes on that side according to the limb side to which each node belongs. The master controller encapsulates the adjusted reporting interval value in a synchronization beacon frame and broadcasts it. After receiving it, each node reports data according to the new interval and re-executes the above process in each cycle.

[0013] Furthermore, in step S7, data preprocessing, frame encapsulation, and double verification specifically involve: Each node stores a new raw data value into its local circular data buffer after collecting it. The oldest value is discarded as soon as a new value is stored. Then, the arithmetic mean of all values ​​in the buffer is calculated as the filtered value for this report. This filtered value is used as the valid value for this collection. For nodes measuring angles and inertia, it is also necessary to continuously collect 10 angle and attitude values ​​and calculate their arithmetic mean as the zero-drift reference value when the node is powered on and initialized and the exoskeleton is in a static upright posture. Subsequent filtered values ​​are subtracted from the corresponding zero-drift reference value to obtain the corrected effective value. The node fills the corrected valid value into the data frame payload area in the order of its own short address, data type code, valid value, timestamp, and gait phase label. Then, it performs cyclic redundancy check on all fields in the payload area to obtain a 16-bit check code A, which is filled into the frame tail. Finally, it performs a byte-by-byte XOR operation on all fields, including check code A, to obtain an 8-bit check code B, which is filled into the frame tail. After receiving a data frame, the master controller first verifies the XOR check code B. If it does not match, the frame is discarded. If it matches, it further verifies the cyclic redundancy check code A. If it matches, the data is parsed. If it does not match, a retransmission request is sent to the source node.

[0014] The technical solution of this invention can achieve the following technical effects: This invention introduces dynamic compensation for wireless signal attenuation caused by joint bending angles, enabling communication link quality assessment to adaptively adjust to real-time changes in joint angles during human walking. This effectively reduces misjudgments and retransmissions caused by signal obstruction due to limb swinging. Simultaneously, by utilizing the symmetrical motion pattern of alternating left and right limb swings during walking, cross-validation and joint correction of the communication quality of the left and right subnets are performed, further improving link reliability under extreme postures. Furthermore, this invention can dynamically adjust the data reporting interval of each node based on gait phase, enabling temporal redistribution of communication resources between the standing and swinging phases, tilting communication bandwidth towards critical phases while maintaining a relatively constant total number of reports. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a flowchart illustrating the Zigbee-based exoskeleton sensor networking method. Figure 2 This is a logic diagram of a Zigbee-based exoskeleton sensor networking method. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The terminology used herein includes any and all combinations of one or more of the associated listed items.

[0018] This invention relates to a Zigbee-based exoskeleton sensor networking method, such as... Figures 1-2As shown, the process includes multiple steps S1 to S7. These steps establish a Zigbee wireless sensor network optimized for the mechanical structure of the exoskeleton, enabling low-latency, high-reliability wireless data communication between the main controller and the sensor nodes distributed at various joints of the exoskeleton. The specific details of each step are as follows: S1: The exoskeleton's master controller scans all Zigbee channels and selects the channel with the least interference to establish the network. Short addresses are assigned to each node based on limb side and joint level depth. The exoskeleton sensor nodes are distributed across the left and right thighs, left and right calves, and left and right feet, requiring support for more than 24 sensor nodes online simultaneously. Regarding channel selection, the 2.4GHz band has many interference sources (such as WiFi and Bluetooth). By scanning all channels and selecting the channel with the least interference to establish the network, the master controller can effectively avoid co-channel interference and ensure communication quality. For address allocation, orderly allocation according to limb side and joint level depth ensures continuous node addresses and clear hierarchical relationships within the same limb subnet, facilitating subsequent topology construction and routing management.

[0019] S2: Construct all nodes into a symmetrical double-subtree topology rooted at the master controller. The left limb forms an independent subtree, and the right limb forms an independent subtree. Both subtrees are rooted at the waist master controller. Data from the left lower limb is transmitted upwards only within the left limb subnet, and data from the right lower limb is transmitted upwards only within the right limb subnet. This avoids the additional latency and power consumption caused by cross-limb jumps and ensures that the data flow is completely consistent with the human mechanical chain.

[0020] S3: As the joint angles of the exoskeleton change continuously during walking, the degree of obstruction of wireless signals by human tissue also changes. In order to accurately assess the actual communication quality of each link, the master controller periodically sends probe frames to each node to collect the received signal strength, link quality indication and end-to-end delay value of each node. At the same time, it obtains the current bending angle value of important joints for subsequent signal compensation and correction.

[0021] S4: The main controller uses the bending angle value to calculate the dynamic attenuation compensation factor of human tissue to wireless signal and compensates and corrects the signal strength to eliminate the additional attenuation caused by joint bending; then it combines the corresponding joint angle of the opposite limb to calculate the symmetry confidence offset of the left and right limbs, and uses the symmetry law of the left and right limb movement during walking to perform secondary correction on the compensated signal strength to reduce the reliability of the measurement results when the gait is abnormal and asymmetrical.

[0022] S5: The master controller calculates the comprehensive communication confidence score for each node using the signal strength after secondary correction. The comprehensive communication confidence score integrates information from three dimensions: signal strength, link quality, and end-to-end delay. A higher score indicates better overall link quality. Based on the score range, the master controller independently selects the most suitable communication mode for each node: direct communication (lowest latency) when signal quality is excellent; forwarding at each level along the parent-child hierarchy when signal quality is moderate (higher reliability); and triggering route repair (finding an alternative path) when signal quality is poor.

[0023] S6: The master controller periodically broadcasts synchronization beacons. Upon receiving the beacons, each node determines its current gait phase based on the resultant force of its plantar pressure and dynamically adjusts its data reporting interval accordingly. During walking, the data update rate requirements differ between the standing phase (foot on the ground bearing weight) and the swinging phase (foot off the ground swinging forward): data changes slowly during the standing phase, requiring no excessively high update rate; the swinging phase, about to land, requires a higher update rate to capture events such as landing impact. This step determines whether the current phase is standing or swinging based on plantar pressure and dynamically adjusts the reporting interval, optimizing communication resource allocation while ensuring control requirements are met.

[0024] S7: To ensure the quality of reported data, each node performs local preprocessing before sending data: Each node performs local moving average filtering and zero drift correction on the collected sensor data, encapsulates the data frame, performs double verification, and then reports it. Moving average filtering eliminates random noise, and zero drift correction eliminates initial sensor deviations, ensuring the reliability of data transmission.

[0025] Preferably, in step S1, the master controller establishes the network as follows: S1.1: After the main control unit is powered on, it scans all 15 Zigbee channels in the 2.4GHz band one by one and measures the average received energy value E on each channel. ch Select E ch The smallest channel is used as the working channel, and network establishment beacon frames are sent on this channel.

[0026] Zigbee has 16 channels (channels 11-26) in the 2.4GHz band. This method uses all 15 available channels (channel 26 may be restricted in some areas; this method includes it in the scanning range for flexible selection). The master controller stays on each channel for a sufficient time to measure the average received energy value, and selects the channel with the lowest energy as the working channel. This can minimize interference from co-channel sources (such as WiFi channels 1, 6, and 11), providing a clean physical layer environment for subsequent communication.

[0027] S1.2: After each node is powered on, it continuously scans the network beacons on the working channel. Upon receiving a beacon, it sends a network access request frame to the master control. The network access request frame carries the node's own pre-configured attribute code. The attribute code consists of three parts: the first part identifies the left or right limb, the second part identifies the joint level to which the node belongs (hip, knee, or ankle), and the third part identifies the node type (routing node or terminal node).

[0028] The attribute code is the core identifier of a node. The master controller makes all decisions regarding subsequent address allocation and topology construction based entirely on the attribute code; therefore, the pre-configuration of the attribute code must be accurate. In actual deployment, the attribute code can be pre-written according to the installation location during node firmware flashing, or it can be manually set during node installation via DIP switches or DIP switches.

[0029] Through the above steps, the master controller can quickly select the optimal channel, and each node can accurately report its installation location and functional role on the exoskeleton, laying the foundation for subsequent topology construction and address allocation.

[0030] Preferably, in step S1, the allocation of the short address specifically involves: S1.3: After receiving the network access request, the master controller reads the limb identifier and level identifier in the attribute code and allocates a short address from the address pool of the corresponding limb subnet: the address pool range of the left limb is 0x0000 to 0x0007, and the address pool range of the right limb is 0x0008 to 0x000F. The allocation is carried out in the order of level depth from shallow to deep, and within the same level, the allocation is carried out in the order of arrival of the network access request.

[0031] In Zigbee networks, a 16-bit short address is the core identifier for node communication. This method divides the address pool according to limb affiliation, with left-limb addresses being consecutive and right-limb addresses being consecutive, which helps the master control quickly determine the limb affiliation of a node by address range. Allocation is done from shallow to deep hierarchy, ensuring that parent nodes obtain addresses first, followed by child nodes, forming a natural numerical order between parent and child addresses, facilitating routing table maintenance and lookup. Addresses 0x0010 to 0x001F are reserved for expansion.

[0032] S1.4: After allocation, the master controller binds the short address with the limb identifier and hierarchical depth in the attribute code and records it in the local address mapping table. It then encapsulates the short address in the network entry response frame and returns it to the corresponding node. After receiving the response frame, the node saves the short address and replies with a network entry confirmation frame to complete the network entry.

[0033] The address mapping table is the core data structure for the master controller to maintain the network topology. The master controller uses the address mapping table to record information such as the short address, limb identification, and hierarchical depth of each node for subsequent routing decisions and link quality tracing. After receiving an inbound response frame, a node saves its short address, and all subsequent communication uses this address, no longer relying on attribute codes.

[0034] S1.5: If the master controller has no available addresses in the address pool of a certain limb subnet, it shall sequentially allocate the next address from the reserved extended address pool 0x0010 to 0x001F and mark the address as the extended address of the limb subnet.

[0035] A reserved extended address pool ensures good system scalability. When a limb needs more sensor nodes (such as adding electromyography sensors, temperature sensors, etc.), the master controller can allocate addresses from the reserved pool to that limb. Marking extended addresses informs the system that these addresses do not belong to the regular hierarchical address level and may require special handling in topology management.

[0036] Through the above steps, each node obtains a unique 16-bit short address, and the master controller establishes a complete address mapping table. All communication above the network layer is based on the short address, without relying on IEEE long addresses, effectively reducing communication overhead.

[0037] Preferably, in step S2, constructing the bisubtree topology specifically involves: S2.1: Mark the master node's topology depth as 0 and its limb side as trunk. The master node, as the root node of the entire network, has a topology depth of 0 and is the final convergence point of all data flows. Marking its limb side as trunk indicates that it does not belong to either the left or right limb subsystem, but is the common root of both the left and right subtrees.

[0038] S2.2: The joint angle sensors and inertial measurement units located on the left and right thighs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 1 for both. The joint angle sensors and inertial measurement units located on the left and right lower legs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 2 for both. The plantar pressure sensors located on the left heel, left toe, right heel, and right toe are labeled as left limb terminal nodes and right limb terminal nodes, respectively, with a topology depth value of 3 for both.

[0039] Topology depth values ​​reflect a node's hierarchical position within the network. A node with a depth of 1 is a direct child of the master node, a node with a depth of 2 is a child of a node with a depth of 1, and a node with a depth of 3 is a child of a node with a depth of 2. Following the mechanical chain structure of an exoskeleton (hip → knee → ankle), the depth relationship naturally corresponds to the physical structure from the torso to the extremities, and the data flow is completely consistent with the human mechanical chain. In sensor configurations, joint angle sensors and inertial measurement units are typically deployed at the hips and knees. These locations are close to the core of the torso and need to undertake both data acquisition and forwarding responsibilities; therefore, they are configured as routing nodes. Foot pressure sensors are located at the extremities and only need to report their own pressure data, without undertaking forwarding tasks for other nodes. Therefore, they are configured as terminal nodes, allowing them to retain only necessary detection and communication functions while disabling other unnecessary functions to reduce power consumption and achieve longer battery life.

[0040] S2.3: The connection relationship of the parent-child link in the topology is established step by step according to the topology depth: the node with a depth of 1 only initiates network entry to the master control with a depth of 0 and establishes a parent-child link; the node with a depth of 2 only initiates network entry to the node with a depth of 1 on the same limb side; the terminal node with a depth of 3 only initiates network entry to the node with a depth of 2 on the same limb side; no direct communication link is established between the left and right limbs.

[0041] This rule clarifies the direction of parent-child link establishment: when a left thigh routing node with depth 1 joins the network, it can only choose the master node as its parent node; when a left calf routing node with depth 2 joins the network, it can only choose a left thigh routing node on the same side (left limb) with depth 1 as its parent node. This ensures that the left and right limbs each form independent subtrees, and that the data flow is always a unidirectional convergence direction from the limb tip to the torso. Direct communication links are not established between the left and right limbs, avoiding the additional latency and network topology complexity caused by cross-limb jumps.

[0042] Preferably, in step S3, the data acquisition specifically involves: S3.1: The master controller sends probe frames to each node sequentially at a set period (1 second in this embodiment). The probe frame header contains a timestamp of the sending time, used to measure the end-to-end delay.

[0043] S3.2: After receiving a probe frame, each node measures the received signal strength (RSSI) of that frame. raw,i Link Quality Indicator (LQI) i The system records the received timestamp and includes it in the reply frame. If the routing node is located at the left thigh, right thigh, left calf, or right calf, it also needs to read the bending angle value θ at its respective joint. i The specific correspondence is as follows: the left thigh node reads θ. hip,L The right thigh node reads θ. hip,R Read θ from the left calf node knee,LRead θ from the right calf node knee,R Then, fill in the angle value into the response frame.

[0044] S3.3: After receiving the reply frames from each node, the master controller subtracts the transmission timestamp in the probe frame from the current time to obtain the end-to-end delay D of that node. i .

[0045] Preferably, in step S4, calculating the dynamic attenuation compensation factor and compensating for the signal strength specifically involves: S4.1: Calculate the dynamic attenuation compensation factor (DAF) of the i-th node. i : DAF i =1+[sin(θ i [×π / 180)] 2 ×λ; In the formula, i is the node number, and hip,L, hip,R, knee,L and knee,R are taken as nodes at the left thigh, right thigh, left calf and right calf, respectively. θ i θ represents the current bending angle value corresponding to the i-th node, in degrees; θ represents the bending angle value of nodes at different positions. i The corresponding physical meaning is: θ of the left thigh node hip,L This indicates the angle between the left thigh and the torso (waist), i.e., the rotation of the left hip joint; θ of the right thigh node hip,R This indicates the angle between the right thigh and the torso (waist), i.e., the rotation of the right hip joint; θ of the left calf node knee,L This indicates the angle between the left lower leg and the left thigh, i.e., the rotation of the left knee; θ of the right lower leg node knee,R This indicates the angle between the right lower leg and the right thigh, i.e., the rotation of the right knee; λ is the human tissue occlusion coefficient, a fixed value of 0.18, which is dimensionless. λ=0.18 indicates that the maximum additional attenuation is approximately 18%, which is the average attenuation coefficient of human tissue to 2.4GHz signals, as determined experimentally.

[0046] The principle behind this formula is as follows: Taking the left thigh as an example, θ hip,L When the angle is 0°, the left thigh hangs vertically below the torso. After the signal is emitted from the waist, it spreads smoothly along the surface of the vertical thigh. The signal path is not blocked at the root of the thigh, so the attenuation is minimal. θ hip,LWhen the angle is 90°, the left thigh is raised forward to a horizontal position. At this time, the root of the thigh is directly in front of the torso, which blocks the direct path between the main control and the thigh node the most tightly and attenuates the most. θ hip,L At 180°, the left thigh folds completely backward or forward. Of course, this will not occur during normal walking of an exoskeleton (and usually rarely exceeds 90°). But even if it does occur, the thigh will close to the torso again, and the root of the thigh will no longer be vertical in the middle of the path as it is at 90°. The signal can bypass the torso surface from the side and attenuate and drop back.

[0047] For the lower leg node, similar to the thigh, when the lower leg bends at a right angle to the thigh, the lower leg itself (from the perspective of the antenna of the lower leg node itself) curves up, forming a huge blocking surface at the knee joint perpendicular to the original signal propagation direction. The lower leg blocks the direct path between the thigh node and the lower leg node most completely, resulting in the greatest attenuation. The variation of this blocking capability with the bending angle θ is approximately the same as the variation of a sine function, therefore a sine function is used for fitting calculation in the formula.

[0048] S4.2: Calculate the equivalent signal strength of the i-th node: RSSI corr,i =RSSI raw,i +10×log 10 (DAF i ); In the formula, RSSI raw,i The original received signal strength of the i-th node measured by the master controller, in dBm; DAF i Let be the dynamic attenuation compensation factor for the i-th node, which is dimensionless.

[0049] The principle behind this formula is as follows: Bending of the joint causes a decrease in the measured signal, and the original RSSI... raw,i The data already includes the additional attenuation caused by joint flexion, which is 10 × log 10 (1 / DAF i ) dB. To compensate for this attenuation, it is necessary to adjust the RSSI. raw,i Add 10×log 10 (DAF i The compensation value of ) dB will correspondingly increase the signal strength, preventing the master controller from being affected by spurious fluctuations caused by joint bending when evaluating the communication quality of the node. Taking θ=90° as an example, DAF i =1.18, compensation amount is 10×log 10 (1.18) = 0.72dB, meaning that when the joint is bent to 90°, the measured signal strength is underestimated by about 0.72dB. This compensation correction adds back the underestimate.

[0050] Through the above compensation and correction, the signal strength values ​​of each node's communication are uniformly corrected to the reference state of the joint being fully extended, eliminating the periodic interference of joint angle changes on the signal strength measurement values.

[0051] Preferably, in step S4, the calculation of the confidence offset of the left and right limbs and the secondary correction of the compensated signal strength are specifically as follows: S4.3: The main controller reads the hip joint angle θ of the left and right limbs at the same moment. hip,L and θ hip,R and knee joint angle θ knee,L and θ knee,R All units are degrees. Calculate the confidence offset Δ for left and right limb symmetry. sym : Δ sym =(θ hip,L -θ hip,R )-(θ knee,L -θ knee,R ).

[0052] The principle behind this formula is as follows: In a normal symmetrical gait, the angle difference between the left and right hips is equal to or approximately equal to the angle difference between the left and right knees, and the degree of posture change on both sides of the limbs is basically symmetrical. Δ sym A value close to zero indicates that the degree of change in posture between the left and right limbs is basically symmetrical. If Δ sym If the absolute value deviates significantly from zero (e.g., more than 10°), it indicates an asymmetry in the degree of posture change between the left and right sides of the limbs, such as limping, signs of impending tripping, or uneven load distribution on the exoskeleton. In this case, the asymmetry in the degree of posture change may cause unpredictable changes in the wireless signal propagation environment, resulting in an imbalance in the signal propagation environment on both sides and a decrease in the reliability of signal strength measurements at each node.

[0053] It is important to note that Δ sym It can only reflect whether the degree of change in the posture of the left and right limbs is symmetrical and whether the signal propagation environment is consistent. It cannot directly determine whether the gait is in a normal state. For example, an abnormal gait, such as both legs bending forward, will also cause Δ to be affected. sym The value is close to zero, but at this point, the changes in limb posture on both sides are basically symmetrical, and the signal propagation environment on both sides is balanced, so no weight reduction is needed in subsequent calculations. If you want to determine whether the gait is abnormal, you need to design an additional judgment method in the master control, such as identifying the hip joint angle θ. hip,L and θ hip,R Methods such as whether the steering is consistent.

[0054] S4.4: Calculate the symmetric confidence correction factor (MCF): MCF=exp(-|Δ sym | / δ); In the formula, |Δ sym | represents the absolute value of the confidence offset for left and right limb symmetry, in degrees; δ is the symmetry attenuation constant, which is a fixed value of 30, and the unit is degrees. MCF is a symmetric confidence correction factor, which is dimensionless.

[0055] The principle of this formula is: when Δ sym When |Δ = 0, the left and right limbs are completely symmetrical, and no weighting is required for the signal strength of each node; the MCF takes the maximum value of 1. sym As the value gradually increases from 0, the gait asymmetry worsens, the overall reliability of signal strength measurements at all nodes decreases, and the MCF decreases monotonically in an exponential manner. δ=30° controls the attenuation rate: |Δ sym When |=30°, MCF=e -1 ≈0.37 indicates that when the gait asymmetry reaches 30°, the reliability of the signal strength measurements at each node decreases to about 37% of that in the symmetrical state. The closer the MCF is to 1, the better the current gait symmetry and the higher the overall reliability of the signal strength measurements at each node.

[0056] S4.5: Calculate the final signal strength value RSSI of the i-th node. joint,i : .

[0057] The principle behind this formula is as follows: Since signal strength is physically characterized by linear power (mW), while in engineering it is commonly expressed in dBm as its logarithmic form, the RSSI is first adjusted during calibration. corr,i Convert from dBm to linear power value Multiplying by the MCF and then taking the logarithm back to the dBm domain ensures dimensional consistency and physical correctness. The MCF is used to perform overall weighting of signal strength measurements across all nodes. RSSI joint,i It integrates joint bending occlusion compensation and left and right limb symmetry confidence correction of each node. The larger the value, the higher the equivalent communication quality of the link in the current gait state.

[0058] After joint flexion compensation correction and left-right limb symmetry confidence correction, the signal strength values ​​RSSI of each node are... joint,i Spurious fluctuations caused by joint angle changes have been eliminated, and gait symmetry confidence has been weighted. Next, signal strength, link quality indication, and end-to-end delay need to be fused to calculate a comprehensive communication confidence score. Preferably, in step S5, the calculation of the comprehensive communication confidence score for each node is as follows: S5.1: Calculate the overall communication confidence score Φ of the i-th node. i : Φ i=[1+(RSSI joint,i -R thr ) / |R thr |]×(LQI i / L thr )×exp(-D i / D thr )×Γ i ; In the formula, LQI i The link quality indicator value of the i-th node measured by the master controller is dimensionless and is directly provided by the Zigbee protocol stack (existing technical parameters). D i Let be the end-to-end delay of the i-th node, in milliseconds. R thr The signal strength reference threshold is a fixed value of -65dBm; L thr This is a baseline threshold for indicating link quality, fixed at 80 (the good link quality threshold defined by the Zigbee protocol stack). D thr The delay baseline threshold is fixed at 20ms. Γ i Let be the historical confidence inheritance factor of the i-th node, with an initial value of 1.0 and dimensionless.

[0059] The principle of this formula is: the first term 1 + (RSSI) joint,i -R thr ) / |R thr |In the middle, when RSSI joint,i =R thr This item is 1 when RSSI = -65dBm; joint,i When the signal strength is better than -65dBm (e.g., -60dBm), this value is greater than 1, indicating sufficient signal strength; when RSSI joint,i When the signal strength is below -65dBm (e.g., -75dBm), this value is less than 1, indicating insufficient signal strength. This design avoids directly using RSSI. i / R thr The non-monotonicity problem arises because the denominator is negative. The second term, LQI... i / L thr This reflects the degree to which digital communication quality is relative to a good threshold; the higher the LQI, the closer this term is to or exceeds 1. The third term exp(-D) i / D thr The penalty for delay on confidence is described in exponential form, D. i When =0ms, this item is 1, D i When =20ms, this item is e. -1 ≈0.37, the greater the delay, the heavier the penalty. The fourth term Γ iBy introducing historical states to suppress drastic jumps in confidence, the memory effect of wireless link quality is simulated. The final obtained Φ i The larger the value, the better the overall communication quality of the i-th link, the stronger the signal, the better the digital communication quality, the lower the latency, and the more stable the historical status.

[0060] S5.2: Γ i The update rule is: each time Φ is calculated i Use the Γ updated in the previous cycle i The value is included in the calculation for this period; after the calculation is completed, if Φ for this period... i If it is below 0.4, then Γ i Updated to Γ i ×0.98; if Φ in this period i Higher than 0.7 and the previous period Φ i If it is also higher than 0.7, then Γ i Updated to Γ i ×1.01; other cases Γ i Remain unchanged. Γ i The value range is limited to [0.3, 1.0]. The master controller independently maintains its own Γ for each node. i Value and Φ of the previous period i Value (denoted as Φprev) i (Initial value: 0.5).

[0061] Γ i The update rule gives the confidence score a slow rise and fast fall characteristic: when the link quality remains good, Γ i A slow increase (1% each time) represents a gradual accumulation of confidence in the system's link stability; when the link quality suddenly drops (Φ... i When <0.4), Γ i A rapid decrease (-2% each time) indicates a sharp deterioration in the system's rapid response. This asymmetric update rate can avoid frequent route switching in unstable environments, while responding quickly when the link truly deteriorates.

[0062] S5.3: Based on the Φ of each node i Each value makes its own routing decision independently: When Φ i A value ≥0.75 indicates that the communication link between the node and the master controller is of excellent quality. In this case, the master controller selects the direct communication mode for the node, that is, the node (source node) sends the data frame directly to the target node (master controller) without passing through any intermediate nodes for forwarding, resulting in the lowest latency (only one hop) and the lowest power consumption (no additional forwarding required).

[0063] When 0.45≤Φ iA value <0.75 indicates that the quality of the direct communication link between the node and the master controller is moderate. Direct communication may not be stable enough, but reliable transmission is still possible through hierarchical forwarding. In this case, the master controller selects the cluster tree routing mode for the node, that is, the source node forwards data hierarchically along the parent-child hierarchy: the terminal node sends the data frame to its parent routing node, the parent routing node receives it and then forwards it to its own parent node, and so on until it reaches the master controller.

[0064] For example, the data flow of the left foot terminal node is: left foot terminal node → left calf routing node (parent node) → left thigh routing node (parent node) → master control. In this mode, the left foot terminal node is the source node, and the data reaches the target node (master control) after being forwarded by two routing nodes.

[0065] Cluster tree routing mode utilizes pre-established fixed parent-child links. Although the increased number of hops leads to a slight increase in latency, the communication distance of each hop is short and the link reliability is high, making it suitable for nodes such as the lower leg and foot that are far from the main control or severely blocked by joints.

[0066] When Φ i When the value is less than 0.45, it indicates that the current communication link quality of the node is severely degraded. If the current path is continued to be used (whether it is direct communication or cluster tree routing along the existing parent-child links), the data packet loss rate will increase significantly. At this time, the master controller selects the route repair mode for the node. The master controller traverses all the candidate neighboring nodes of the node, calculates the Φ value of each candidate node, selects the one with the highest Φ as the new path, generates a route repair command and sends it to the source node. The source node updates its local routing table and reports data according to the new path.

[0067] In this context, the source node refers to the node currently reporting data, i.e., the initiator of the data. Any sensor node (including routing nodes and terminal nodes) acts as a source node when it needs to report data to the master controller. For example, when the routing node of the left thigh needs to report its joint angle data, it is the current source node for communication. Within a synchronization cycle, each node with a data reporting requirement independently performs communication as a source node.

[0068] The target node refers to the final destination of the data report, namely the lumbar master controller. In the exoskeleton system described in this paper, all sensor data must ultimately converge to the master controller for processing and control decisions. Therefore, the target node is fixed to the master controller and remains unchanged in all communication modes.

[0069] Alternate neighboring nodes refer to other nodes that have a parent-child relationship with the source node, including the source node's parent and child nodes. When the quality of the source node's current communication link deteriorates, the master node can select one of the alternative neighboring nodes as a new relay node to help the source node forward data to the master node. For example, the parent node of the left thigh routing node is the master node, and the child node is the left calf routing node; therefore, its alternative neighboring nodes include the master node (parent node) and the left calf routing node (child node). It is important to note that although the master node is the parent node of the left thigh node, in route repair mode, the master node as an alternative neighboring node is meaningless (because the master node is already the final destination). Therefore, the actual alternative for route repair is the child node (such as the left calf node).

[0070] In the preferred step S6, the dynamic adjustment of gait phase determination and reporting interval is specifically as follows: S6.1: Each foot terminal node collects pressure values ​​at four points: left heel, left toe, right heel, and right toe. The heel pressure value and toe pressure value on the same limb side are summed to obtain the resultant plantar pressure value of that limb side, thereby indirectly reflecting the weight-bearing status of that limb.

[0071] S6.2: Before broadcasting the synchronization beacon, the master controller acquires the combined plantar pressure values ​​of both sides once, and takes the larger one as the effective plantar pressure value F for that limb side. eff Then the main controller according to F eff Determine the current gait phase on this side of the limb: Set the judgment threshold F thr When F eff ≥F thr When it is determined that the side of the limb is in the standing phase, when F eff <F thr It is determined that the device is in the oscillation phase.

[0072] F thr The specific method for determining the threshold is as follows: each limb side is calibrated independently. The maximum plantar pressure value is measured at the initial stage of the standing phase on that side (i.e., the instant the foot just touches the ground), and 50% of that value is taken as the threshold. This adaptive calibration method can adapt to wearers of different weights and different pace speeds, avoiding misjudgments caused by using a fixed threshold.

[0073] S6.3: The master controller determines the reporting interval adjustment coefficient ρ for each node based on the gait phase determination results: ρ is 1.2 for standing phase nodes and 0.8 for swinging phase nodes. The master controller carries the adjusted reporting interval I corresponding to each node in the synchronization beacon frame. adi =I base ×ρ, where I base As a reference reporting interval, the I of the inertial measurement unit and the joint angle sensor in the sensor. baseThe I of the plantar pressure sensor is 20ms. base It takes 30ms.

[0074] The reporting intervals of all nodes on each side were adjusted. The reporting interval for the standing phase nodes was increased by 20% (from 20ms to 24ms, and from 30ms to 36ms). This is because the data for the standing phase changes slowly, and reducing the update rate does not affect control accuracy, while also saving power consumption and channel resources. The reporting interval for the swinging phase nodes was shortened by 20% (from 20ms to 16ms, and from 30ms to 24ms). This is because a higher data update rate is needed when the swinging phase is about to hit the ground to capture transient events such as the impact of landing.

[0075] S6.5: After receiving the synchronization beacon, each node parses the I allocated to it from the beacon frame. adi The value is used to start a timer, and the sensor data collected in this cycle is reported immediately when the timer expires. Each node recalculates and updates its local I value upon receiving a synchronization beacon. adi Each time a synchronization beacon is received, each node reacquires the adjusted interval value and updates its local timer, enabling real-time dynamic adjustment of the reporting interval according to the gait phase.

[0076] This step identifies gait phases through plantar pressure and dynamically adjusts the reporting interval to optimize communication resource allocation while ensuring control requirements are met. During the standing phase, the foot bears weight on the ground, and the joint angle and plantar pressure change slowly, so a high data update rate is not required. During the swinging phase, the foot swings forward off the ground, and a higher update rate is needed to capture events such as landing impact as the foot is about to land.

[0077] In the preferred step S7, data preprocessing, frame encapsulation, and double verification specifically include: S7.1: After each node collects a new raw data value, it stores it in a local circular data buffer with a length of 5. Each time a new value is stored, the oldest value is discarded. Then, the arithmetic mean of all 5 values ​​in the buffer is calculated as the filtered value for this report, and this filtered value is used as the valid value for this collection.

[0078] For nodes measuring angles and inertia (i.e., joint angle sensors and inertial measurement units), it is necessary to continuously collect 10 angle and attitude values ​​while the node is powered on and the exoskeleton is in a static upright posture, and calculate the arithmetic mean of each value as the zero-drift reference value, which is then stored in a local non-volatile register. After each subsequent filtering, the filtered value is subtracted from the corresponding zero-drift reference value to obtain the corrected effective value.

[0079] Zero drift is an inherent systematic error of sensors; even without physical input, a sensor may output a non-zero value. This method takes 10 samples in a static, upright position and averages them as the zero drift baseline. This effectively eliminates random errors at individual sampling points, obtaining a more accurate zero drift estimate. Subsequent filtered values ​​are subtracted from the corresponding zero drift baseline. After subtracting the baseline, subsequent measurements eliminate systematic bias, improving accuracy.

[0080] S7.2: The node fills the corrected effective value into the data frame payload area in the order of its own short address, data type code, effective value, timestamp, and gait phase label. The short address is used to identify the data source, the data type code is used to distinguish different sensor types, the effective value is the core data, the timestamp is used to ensure the timeliness of the data, and the gait phase label provides context information for subsequent control algorithms.

[0081] S7.3: After filling is completed, perform Cyclic Redundancy Check (CRC) on all fields in the payload area from the short address to the gait phase tag to obtain a 16-bit check code A, which is then filled into the first check field of the frame tail. Subsequently, perform an XOR operation on all fields, including check code A, byte by byte to obtain an 8-bit check code B, which is then filled into the second check field of the frame tail.

[0082] CRC checksum (check code A) can detect multi-bit errors, but it requires a large amount of computation and is highly reliable. XOR checksum (check code B) is simple to calculate and can quickly detect single-bit errors. The two checksum methods are complementary. XOR checksum serves as a fast pre-check; if B does not match, the frame is discarded directly, avoiding time-consuming CRC calculations on obviously corrupted frames. CRC checksum serves as a precise final check, ensuring data integrity.

[0083] S7.5: After receiving a data frame, the master controller first calculates the XOR check value of the entire frame and compares it with the check code B. If they do not match, the frame is discarded; if they match, the cyclic redundancy check value is further calculated and compared with the check code A. If they match, the payload data is parsed; otherwise, a retransmission request is sent to the source node, effectively reducing unnecessary computational overhead. CRC check is only performed when the XOR check passes, and retransmission is only triggered when the CRC check fails, reducing unnecessary network retransmission traffic.

[0084] Taking a certain model of exoskeleton walking on flat ground as an example, the steps of this method will be further explained. The main control unit of the exoskeleton is deployed in the waist, with one routing node (containing joint angle sensors and inertial measurement units) deployed in each of the left and right thighs and calves, and one plantar pressure terminal node deployed in each of the left and right heels and toes, for a total of 1 main control unit, 4 routing nodes, and 4 terminal nodes. The specific process is as follows: Network initialization phase: After scanning 15 Zigbee channels, the master controller found that channel 11 had the lowest average energy (-89dBm), and selected channel 11 to establish the network. When the left thigh node joined the network, it carried the attribute code "left limb-hip layer-routing," and the master controller assigned it a short address of 0x0000; the left lower leg node (left limb-knee layer-routing) was assigned 0x0001; the right thigh node (right limb-hip layer-routing) was assigned 0x0008; and the right lower leg node (right limb-knee layer-routing) was assigned 0x0009. Each foot terminal joined the network sequentially according to its hierarchical level, with topology depths of 1, 2, and 3, forming a complete left-right symmetrical bi-subtree. After the network was established and running stably, it entered the periodic probing and routing decision-making phase.

[0085] Link quality detection and compensation / correction phase: The probe data reported by four routing nodes at a certain moment are as follows: Number i=1 (left thigh node), θ1=θ hip,L =35°, RSSI raw,1 =-72dBm, LQI1=76, D1=15ms; Number i=2 (right thigh node), θ2=θ hip,R =40°, RSSI raw,2 =-70dBm, LQI2=78, D2=14ms; Number i=3 (left lower leg node), θ3=θ knee,L =12°, RSSI raw,3 =-68dBm, LQI3=80, D3=12ms; Number i=4 (right lower leg node), θ4=θ knee,R =18°, RSSI raw,4 =-69dBm, LQI4=79, D4=13ms.

[0086] RSSI raw =-72dBm to -68dBm is the typical received signal strength range for 2.4GHz wireless communication on the human body surface; LQI=76~80 is the measured good link quality range of the Zigbee protocol stack (LQI≥70 indicates good link quality); D=12~15ms is the typical end-to-end delay for single-hop or multi-hop Zigbee.

[0087] Dynamic attenuation compensation factor (DAF) i calculate: Number i=1 (left thigh, θ1=35°): DAF1 = 1 + [sin(35° × π / 180)] 2 ×0.18 = 1 + (0.574) 2 ×0.18=1.059; RSSI corr,1 =RSSI raw,1 +10×log 10 (DAF1) = -72 + 10 × log 10 (1.059) = -72 + 0.25 = -71.75 dBm; Number i=2 (right thigh, θ2=40°): DAF² = 1 + [sin(40° × π / 180)] 2 ×0.18 = 1 + (0.643) 2 ×0.18=1.074; RSSI corr,2 =-70+10×log 10 (1.074) = -70 + 0.31 = -69.69 dBm; Number i=3 (left calf, θ3=12°): DAF3 = 1 + [sin(12°×π / 180)] 2 ×0.18 = 1 + (0.208) 2 ×0.18=1.008; RSSI corr,3 =-68+10×log 10 (1.008) = -68 + 0.03 = -67.97 dBm; Number i=4 (right calf, θ4=18°): DAF4 = 1 + [sin(18° × π / 180)] 2 ×0.18 = 1 + (0.309) 2 ×0.18=1.017; RSSI corr,4 =-69+10×log 10 (1.017) = -69 + 0.07 = -68.93 dBm.

[0088] Symmetric confidence offset and secondary correction: Confidence offset for left and right limbs: Δ sym =(θ1-θ2)-(θ3-θ4)=(35-40)-(12-18)=(-5)-(-6)=1°; Δ sym =1° is close to zero, indicating that the current gait is basically symmetrical between the left and right limbs, and the measurement results are highly reliable.

[0089] Symmetric confidence correction factor: MCF=exp(-|Δ sym| / δ)=exp(-|1| / 30)=exp(-0.0333)=0.967.

[0090] Final signal strength after secondary correction at each node: Number i=1 (left thigh): {-71.75 / 10}×0.967)=-71.90dBm; Number i=2 (right thigh): ; Number i=3 (left calf): ; Number i=4 (right calf): .

[0091] Comprehensive communication confidence score Φ i calculate: Each node Γ i Currently, all are 1.0. Substitute them into the formula to calculate Φ. i : Number i=1 (left thigh): Φ1=[1+(RSSI joint,1 -R thr ) / |R thr |]×(LQI1 / L thr )×exp(-D1 / D thr )×Γ1=[1+(-71.90+65) / 65]×(76 / 80)×exp(-15 / 20)×1.0=0.894×0.95×0.472=0.401; Number i=2 (right thigh): Φ2=[1+(-69.84+65) / 65]×(78 / 80)×exp(-14 / 20)×1.0=0.926×0.975×0.497=0.449; Number i=3 (left calf): Φ3=[1+(-68.12+65) / 65]×(80 / 80)×exp(-12 / 20)×1.0=0.952×1.0×0.549=0.523; Number i=4 (right calf): Φ4=[1+(-69.08+65) / 65]×(79 / 80)×exp(-13 / 20)×1.0=0.937×0.988×0.522=0.483.

[0092] Based on routing decision logic: Number i=1 (left thigh): Φ1=0.401<0.45→ Trigger route repair; Number i=2 (right thigh): Φ2=0.449<0.45→ Trigger route repair; Number i=3 (left calf): Φ3=0.523∈[0.45, 0.75)→cluster tree routing mode; Number i=4 (right calf): Φ4=0.483∈[0.45,0.75)→cluster tree routing mode.

[0093] Route repair execution verification: For the node i=1 (left thigh node, source node) that triggers route repair, the target node is fixed as the waist master node. The master node traverses its candidate neighbor node (left calf node, number i=3, its child node) and measures the RSSI of the candidate node. raw =-66.5dBm, LQI=82, D=11ms. The joint angle of this alternative link is θ3=12°. DAF = 1 + [sin(12° × π / 180)] 2 ×0.18=1.008; RSSI corr =-66.5+10×log 10 (1.008)=-66.5+0.03=-66.47dBm; ; Φ bak =[1+(-66.62+65) / 65]×(82 / 80)×exp(-11 / 20)×1.0=0.975×1.025×0.576=0.576; Φ bak =0.576>0.45, the master switch will switch the left thigh node to the alternative adjacent node (left calf node), and the new path will be "left thigh node → left calf node → master". The route repair was successfully executed.

[0094] For the node i=2 (right thigh node, source node) that triggers route repair, the master traverses its candidate neighbor node (right calf node, number i=4) and measures the RSSI of the candidate node. raw =-67.1dBm, LQI=80, D=12ms: θ4 = 18°, DAF = 1.017, RSSI corr =-67.1+0.07=-67.03dBm; ; Φ bak=[1+(-67.18+65) / 65]×(80 / 80)×exp(-12 / 20)×1.0=0.966×1.0×0.549=0.530; Φ bak =0.530>0.45, the master switcher switches the right thigh node to the alternative adjacent node (right calf node). Route repair executed successfully.

[0095] Simulation verification of DAF compensation effect: To further quantitatively evaluate the effect of DAF compensation on signal stability, simulations were performed within a walking gait cycle (1 second, with θ varying sinusoidally from 0° to 70°). RSSI raw The RSSI decreases linearly with increasing angle, with a decay slope of -0.015 dB / °, meaning that at θ=0° the RSSI... raw RSSI = -71dBm, θ = 70° raw = -72.05dBm. This slope is based on the experimentally determined signal attenuation rate when a human joint is bent.

[0096] The results of the before-and-after compensation comparison at 11 equally spaced angles are as follows: At θ=0°, DAF=1.000, compensation amount=0dB, RSSI raw =-71.000dBm, RSSI corr =-71.000dBm; At θ = 7°, DAF = 1.003, compensation = 0.012 dB, RSSI raw =-71.105dBm, RSSI corr =-71.093dBm; At θ = 14°, DAF = 1.011, compensation = 0.045 dB, RSSI raw =-71.210dBm, RSSI corr =-71.165dBm; At θ = 21°, DAF = 1.023, compensation = 0.099 dB, RSSI raw =-71.315dBm, RSSI corr =-71.216dBm; At θ = 28°, DAF = 1.040, compensation = 0.169 dB, RSSI raw =-71.420dBm, RSSI corr =-71.251dBm; At θ = 35°, DAF = 1.059, compensation = 0.250 dB, RSSI raw =-71.525dBm, RSSIcorr =-71.275dBm; At θ = 42°, DAF = 1.081, compensation = 0.337 dB, RSSI raw =-71.630dBm, RSSI corr =-71.293dBm; At θ = 49°, DAF = 1.103, compensation = 0.424 dB, RSSI raw =-71.735dBm, RSSI corr =-71.311dBm; At θ = 56°, DAF = 1.124, compensation = 0.506 dB, RSSI raw =-71.840dBm, RSSI corr =-71.334dBm; At θ = 63°, DAF = 1.143, compensation = 0.580 dB, RSSI raw =-71.945dBm, RSSI corr =-71.365dBm; At θ = 70°, DAF = 1.159, compensation = 0.640 dB, RSSI raw =-72.050dBm, RSSI corr =-71.410dBm.

[0097] RSSI before compensation raw The peak-to-peak variation was -71.000 dBm to -72.050 dBm, or 1.050 dB. The compensated RSSI... corr The peak-to-peak value variation was -71.000 dBm to -71.410 dBm, or 0.410 dB. The compensation reduced the signal fluctuation amplitude from 1.050 dB to 0.410 dB, a reduction of 61%, indicating that DAF compensation effectively eliminated the periodic disturbance of the signal strength measurement value caused by the change of joint angle of each node.

Claims

1. A method for networking exoskeleton sensors based on Zigbee, characterized in that, include: S1: The master controller scans all Zigbee channels and selects the channel with the least interference to establish the network, and assigns short addresses to each node according to the limb side and joint level depth; S2: Construct a left-right symmetrical twin-tree topology with the master node as the root of all nodes; S3: The master controller periodically sends probe frames to each node to collect the received signal strength, link quality indication and end-to-end delay of each node, and at the same time obtains the current bending angle value of important joints; S4: The main controller uses the bending angle value to calculate the dynamic attenuation compensation factor of human tissue for wireless signal and compensates and corrects the signal strength. Then, it combines the corresponding joint angle of the opposite limb to calculate the symmetrical confidence offset of the left and right limbs and performs secondary correction on the compensated signal strength. S5: The master controller uses the signal strength after secondary correction to calculate the comprehensive communication confidence score of each node, and selects direct communication, cluster tree forwarding or triggering route repair based on the score; S6: The master controller periodically broadcasts a synchronization beacon. After receiving the beacon, each node determines its current gait phase based on its own plantar pressure resultant force value and dynamically adjusts its data reporting interval according to the phase. S7: Each node performs local moving average filtering and zero drift correction on the collected sensor data, encapsulates the data frame, performs double verification, and then reports it.

2. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S1, establishing the network specifically involves: After the main controller is powered on, it scans all Zigbee channels in the 2.4GHz band one by one and measures the average received energy value E on each channel. ch Select E ch The smallest channel is used as the working channel, and network establishment beacon frames are sent on this channel. After powering on, each node continuously scans the network beacons on the working channel. Upon receiving a beacon, it sends a network access request frame to the master controller. The network access request frame carries the node's own pre-configured attribute code, which consists of three parts: the first part identifies the left or right limb, the second part identifies the joint level to which the node belongs, and the third part identifies the node type.

3. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S1, the allocation of a short address specifically involves: After receiving the network access request, the master controller reads the limb identifier and level identifier in the attribute code and allocates a short address from the address pool of the corresponding limb subnet: the address pool range of the left limb is 0x0000 to 0x0007, and the address pool range of the right limb is 0x0008 to 0x000F. The allocation is carried out in the order of level depth from shallow to deep, and within the same level, the allocation is carried out in the order of arrival of the network access request. After allocation, the master controller records the short address and the limb identifier and hierarchical depth binding in the attribute code in the local address mapping table, and encapsulates the short address in the network access response frame and returns it to the corresponding node. After receiving the response frame, the node saves the short address and replies with a network access confirmation frame to complete the network access. If the master controller has no available addresses in the address pool of a certain limb subnet, it will sequentially allocate the next address from the reserved extended address pool 0x0010 to 0x001F and mark the address as the extended address of that limb subnet.

4. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S2, the construction of the twin-tree topology is specifically as follows: The topology depth value of the master controller is marked as 0, and the limb side is marked as trunk; The sensors located on the left and right thighs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 1 for both. The sensors located on the left and right lower legs are labeled as left limb routing nodes and right limb routing nodes, respectively, with a topology depth value of 2 for both. The sensors located on the left heel, left toe, right heel, and right toe are labeled as left limb terminal nodes and right limb terminal nodes, respectively, with a topology depth value of 3 for both. The connection relationship between parent and child links is established step by step according to the topology depth; no direct communication link is established between the left and right limbs.

5. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S3, the specific steps for obtaining data are as follows: The master controller sends probe frames to each node sequentially at a set period. After receiving a probe frame, each node measures the received signal strength and link quality indication of the frame and uses the measurement results as a reply frame. If the nodes are located at the left thigh, right thigh, left calf, and right calf, then the joint angle value of that node at that time should also be read and filled into the response frame; then all response frames should be sent back to the master controller. After receiving the reply frames from each node, the master controller subtracts the sending timestamp in the probe frame from the current time to obtain the delay of that node; In step S4, the calculation of the dynamic attenuation compensation factor and the compensation correction of the signal strength are specifically as follows: Calculate the dynamic attenuation compensation factor (DAF) of the i-th node. i : DAF i =1+[sin(θ i ×π / 180)] 2 ×λ; In the formula, i takes the values ​​hip,L, hip,R, knee,L and knee,R, which represent the nodes at the left thigh, right thigh, left calf and right calf, respectively. λ is the human tissue occlusion coefficient; Calculate the equivalent signal strength of the i-th node: RSSI corr,i =RSSI raw,i +10×log 10 (DAF i ); In the formula, RSSI raw,i Let be the received signal strength of the i-th node.

6. The Zigbee-based exoskeleton sensor networking method according to claim 5, characterized in that, In step S4, the calculation of the confidence offset between the left and right limbs and the secondary correction of the compensated signal strength are specifically as follows: The main controller reads the hip joint angle θ of both sides of the body at the same moment. hip,L and θ hip,R and knee joint angle θ knee,L and θ knee,R Calculate the confidence offset Δ between the left and right limbs. sym : D sym =(θ hip,L -θ hip,R )-(θ knee,L -θ knee,R ); Calculate the symmetric confidence correction factor (MCF): MCF=exp(-|Δ sym | / d); In the formula, δ is the symmetry attenuation constant; Calculate the final signal strength value RSSI of the i-th node. joint,i : 。 7. The Zigbee-based exoskeleton sensor networking method according to claim 6, characterized in that, In step S5, the calculation of the comprehensive communication confidence score for each node is as follows: Calculate the overall communication confidence score Φ of the i-th node. i : Φ i =[1+(RSSI joint,i -R thr ) / |R thr |]×(LQI i / L thr )×exp(-D i / D thr )×Γ i ; In the formula, LQI i D is the link quality indicator for the i-th node; i R is the delay of the i-th node; thr L is the signal strength reference threshold; thr D is the baseline threshold for link quality indication; thr Γ is the delay baseline threshold; i is the historical confidence inheritance factor of the i-th node.

8. The Zigbee-based exoskeleton sensor networking method according to claim 7, characterized in that, In step S5, the specific choices for direct communication, cluster tree forwarding, or triggered route repair are as follows: When Φ i When the value is ≥0.75, the direct communication mode is selected, and the source node sends the data frame directly to the target node; When 0.45≤Φ i When the value is less than 0.75, the cluster tree routing mode is selected, and the source node forwards the data level by level along the parent-child hierarchy. When Φ i When the value is less than 0.45, the route repair mode is triggered. The master controller traverses all the candidate adjacent nodes of the node, calculates the Φ value of each candidate node, selects the one with the highest Φ as the new path, and generates a route repair command to send to the source node.

9. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S6, the gait phase determination and dynamic adjustment of the reporting interval are specifically as follows: The pressure values ​​at four points—left heel, left toe, right heel, and right toe—are collected. The heel pressure value and toe pressure value on the same side of the limb are summed to obtain the resultant plantar pressure value on that side of the limb. Before broadcasting the synchronization beacon, the master controller acquires the combined plantar pressure values ​​of both sides and takes the larger one as the effective plantar pressure value F for that limb. eff ; Set the judgment threshold F thr When F eff ≥F thr When it is determined that the side of the limb is in the standing phase, when F eff <F thr The system determines if the device is in the swing phase; the reporting interval for the standing phase node is adjusted to the baseline value × 1.2, and the reporting interval for the swing phase node is adjusted to the baseline value × 0.8; the main controller adjusts the reporting interval of all nodes on that side according to the limb side to which each node belongs. The master controller encapsulates the adjusted reporting interval value in a synchronization beacon frame and broadcasts it. After receiving it, each node reports data according to the new interval and re-executes the above process in each cycle.

10. The Zigbee-based exoskeleton sensor networking method according to claim 1, characterized in that, In step S7, data preprocessing, frame encapsulation, and double verification specifically involve: Each node stores a new raw data value into its local circular data buffer after collecting it. The oldest value is discarded as soon as a new value is stored. Then, the arithmetic mean of all values ​​in the buffer is calculated as the filtered value for this report. This filtered value is used as the valid value for this collection. For nodes measuring angles and inertia, it is also necessary to continuously collect 10 angle and attitude values ​​and calculate their arithmetic mean as the zero-drift reference value when the node is powered on and initialized and the exoskeleton is in a static upright posture. Subsequent filtered values ​​are subtracted from the corresponding zero-drift reference value to obtain the corrected effective value. The node fills the corrected valid value into the data frame payload area in the order of its own short address, data type code, valid value, timestamp, and gait phase label. Then, it performs cyclic redundancy check on all fields in the payload area to obtain a 16-bit check code A, which is filled into the frame tail. Finally, it performs a byte-by-byte XOR operation on all fields, including check code A, to obtain an 8-bit check code B, which is filled into the frame tail. After receiving a data frame, the master controller first verifies the XOR check code B. If it does not match, the frame is discarded. If it matches, it further verifies the cyclic redundancy check code A. If it matches, the data is parsed. If it does not match, a retransmission request is sent to the source node.

Images