A shopping cart adaptive following control method and system based on elderly users

By calculating the relative position data of the shopping cart with elderly users and obstacles, attraction and repulsion potential functions are constructed to dynamically adjust the following control of the shopping cart. This solves the problems of adaptability and coordination in the following control method of shopping cart for elderly users, and improves safety and stability.

CN123067482APending Publication Date: 2026-11-10CHONGQING UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611579308.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-13
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing smart shopping cart follow control methods lack adaptive capabilities, cannot maintain stable following when elderly users' walking speed changes, and have poor coordination between following and obstacle avoidance functions, resulting in insufficient safety and stability.

Method used

By acquiring relative position data between the shopping cart and the elderly user, as well as obstacle data, the resultant force vector is calculated. Combined with the attraction and repulsion potential functions, the following distance and speed are dynamically adjusted to achieve adaptive following control of the shopping cart, avoiding speed jitter caused by mode switching.

Benefits of technology

It improves the adaptability and coordination of shopping cart following, ensuring the safety and smoothness of use for elderly users, and adapting to stable following in different speeds and complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067482A_ABST
    Figure CN123067482A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of intelligent shopping cart control, in particular to a shopping cart adaptive following control method and system based on old users. The method comprises the following steps: obtaining relative position data and a relative change rate of a shopping cart and a target old user, and obstacle position data and a current speed of the shopping cart; calculating a resultant force vector according to the relative position data and the obstacle position data; calculating a user current speed of the target old user according to the relative change rate and the current speed of the shopping cart; determining an expected following distance according to the user current speed, and determining an expected motion direction and an expected motion speed according to the resultant force vector and the expected following distance; and determining a shopping cart speed control instruction according to the expected motion direction and the expected motion speed. The application solves the technical problems of poor adaptability and poor synergy of the shopping cart adaptive following control method for old users in the prior art.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent shopping cart control technology, specifically a shopping cart adaptive following control method and system for elderly users. Background Technology

[0002] As the population ages, the elderly have become the core consumer force in offline shopping malls and supermarkets. However, traditional hand-pushed shopping carts rely on manual pushing, and when fully loaded, the turning and braking resistance increases significantly. Elderly users generally lack the physical strength to push them, making them difficult to push. Furthermore, their handling stability is poor, and they are prone to accidents such as bumps and falls when encountering obstacles due to insufficient reaction time. Traditional shopping carts can no longer meet the needs of the elderly.

[0003] Several intelligent following shopping cart solutions have emerged in the existing technology, but they generally suffer from the following technical defects: First, the following control lacks the ability to adapt to the target's motion state. Existing solutions typically adopt a fixed-parameter following strategy, where the expected following distance and control gain are fixed in the system design stage. When the target's movement speed changes significantly (such as fluctuations in the walking speed of elderly people or frequent stops), the fixed-parameter controller cannot simultaneously meet the dual requirements of "keeping up in time when the speed is high" and "avoiding approaching when the speed is low," leading to following failure problems such as losing track or feeling pressured.

[0004] Secondly, the coordination mechanism between the follow and obstacle avoidance functions has architectural flaws. Existing solutions often design follow and obstacle avoidance as independent modules in series, merging them through manually set priorities or switching thresholds. When the outputs of the two conflict, the system repeatedly switches between modes or performs weighted fusion, resulting in sudden speed changes and motion oscillations, failing to achieve smooth and continuous follow-obstacle avoidance coordinated behavior.

[0005] Therefore, there is an urgent need for an adaptive following control method for shopping carts used by elderly users that is more adaptable and collaborative. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an adaptive following control method and system for shopping carts used by elderly users, which solves the problems of poor adaptability and poor coordination in existing adaptive following control methods for shopping carts used by elderly users.

[0007] To achieve the above objectives, the present invention provides an adaptive following control method for a shopping cart based on elderly users. The method includes: acquiring relative position data and relative rate of change between the shopping cart and the target elderly user, as well as obstacle position data and the current speed of the shopping cart; calculating a resultant force vector based on the relative position data and the obstacle position data; calculating the current speed of the target elderly user based on the relative rate of change and the current speed of the shopping cart; determining a desired following distance based on the current speed of the user, and determining a desired direction of motion and a desired speed of motion based on the resultant force vector and the desired following distance; and determining a shopping cart speed control command based on the desired direction of motion and the desired speed of motion.

[0008] This invention accurately captures the walking dynamics of the elderly by collecting relative information between the user and the shopping cart, obstacle and vehicle speed data, and calculating the distance change rate. It integrates the user's attraction potential and the obstacle's repulsion potential to generate a resultant force vector, avoiding speed fluctuations caused by switching between follow and obstacle avoidance modes. It calculates the actual user's walking speed based on the relative change rate, dynamically matches and adapts the safe following distance for the elderly, and comprehensively solves the motion direction and speed by combining the resultant force, following deviation, and obstacle distance. Multiple smooth constraints are used to output motor control commands, improving the adaptability and coordination of the elderly following the shopping cart in supermarkets.

[0009] Optionally, acquiring the relative position data and relative change rate between the shopping cart and the target elderly user, as well as the obstacle position data and current speed of the shopping cart, includes: acquiring a sequence of relative position data between the shopping cart and the target elderly user over a predetermined time interval, historical relative position data, obstacle position data, and current speed of the shopping cart; determining the maximum permissible change based on the historical relative position data; performing amplitude limiting filtering on the relative position data sequence based on the maximum permissible change to obtain a first data sequence; performing mean filtering on the first data sequence to obtain a second data sequence; extracting relative position data and the relative position data of the previous period from the second data sequence; and calculating the relative change rate based on the relative position data and the relative position data of the previous period.

[0010] This invention collects multi-period relative position sequences and historical data, sets the maximum allowable change in displacement, implements amplitude limiting filtering to remove occlusion and jump noise, and then uses mean filtering to smooth the interference of human body swaying and irregular movements. Based on stable filtered data, it extracts the positions of previous and subsequent periods and calculates the relative change rate, thereby improving the anti-interference capability of UWB positioning and the accuracy of user motion state estimation.

[0011] Optionally, calculating the resultant force vector based on the relative position data and the obstacle position data includes: obtaining a user position vector and an obstacle point set based on the relative position data and the obstacle position data; constructing an attractive potential function based on the user position vector; constructing a repulsive potential function based on the obstacle point set; calculating the negative gradient of the attractive potential function to obtain an attractive force vector; calculating the negative gradient of the repulsive potential function to obtain a repulsive force vector; and vector-adding the attractive force vector and the repulsive force vector to obtain the resultant force vector.

[0012] This invention constructs user attraction potential function and obstacle repulsion potential function separately, obtains the corresponding attraction and repulsion force vectors by calculating the negative gradient of each, and then superimposes the vectors to obtain the resultant force. It unifies and integrates the needs of following and obstacle avoidance, without the need to set a mode switching threshold, thereby improving the continuity of shopping cart movement and the smoothness of use for elderly users.

[0013] Optionally, constructing the attraction potential function based on the user location vector includes: obtaining the attraction gain base and the timeliness status of the ultra-wideband data of the relative location data; determining the confidence level based on the timeliness status of the ultra-wideband data; calculating the attraction gain using the confidence level and the attraction gain base; and constructing the attraction potential function using the user location vector and the attraction gain.

[0014] This invention reads the timeliness status of UWB data and calculates the confidence level, combines the reference attraction gain with real-time modulation of the effective attraction gain, and then uses the user position vector to construct the attraction potential. When the positioning signal is poor, the following force is automatically weakened, thereby improving the stability of the elderly shopping cart operation under signal obstruction conditions.

[0015] Optionally, constructing the repulsion potential function based on the obstacle point set includes: obtaining the equivalent radius of the shopping cart; expanding each obstacle point in the obstacle point set radially outward by the equivalent radius to obtain the expanded obstacle point set; and constructing the repulsion potential function based on the expanded obstacle point set.

[0016] This invention introduces an equivalent radius of the shopping cart to radially expand each obstacle point, and then constructs a repulsive potential function based on the expanded obstacle point set to effectively enlarge the obstacle safety boundary, fully reserve the physical avoidance margin of the vehicle body, and improve the obstacle avoidance safety of elderly people's shopping carts in narrow passage scenarios in supermarkets.

[0017] Optionally, constructing the repulsion potential function based on the expanded obstacle point set includes: obtaining the repulsion gain and spatial influence scale parameter; calculating the obstacle point repulsion potential function components based on the repulsion gain, the spatial influence scale parameter, and the expanded obstacle point set; and superimposing the obstacle point repulsion potential function components to obtain the repulsion potential function.

[0018] This invention configures repulsion gain and spatial influence scale parameters, solves the repulsion potential components corresponding to each obstacle point by point and superimposes them to obtain the total repulsion potential. It can flexibly adjust the avoidance sensitivity and warning distance, and automatically integrate multiple obstacle avoidance needs in multi-obstacle scenarios, thereby improving the global smooth obstacle avoidance capability of shopping carts in complex pedestrian flow shelving environments in supermarkets.

[0019] Optionally, determining the expected following distance based on the user's current speed includes: obtaining the minimum following distance and the reaction time coefficient; and calculating the expected following distance based on the user's current speed, the minimum following distance, and the reaction time coefficient.

[0020] This invention dynamically calculates the appropriate expected following distance by taking the minimum following distance and reaction time coefficient and combining it with the elderly user's real-time walking speed. At low speeds, the distance is reduced to avoid losing the user due to excessive distance, and at high speeds, the safety buffer is increased, thereby improving the following adaptability and walking safety of elderly users at different paces.

[0021] Optionally, determining the desired motion direction and desired motion speed based on the resultant force vector and the desired following distance includes: extracting the vector direction from the resultant force vector to obtain the desired motion direction; calculating the distance-corrected speed based on the relative position data and the desired following distance; obtaining the equivalent damping coefficient, a preset angle range, and a preset proximity threshold; calculating the resultant force modulus based on the resultant force vector, and calculating the base speed based on the resultant force modulus and the equivalent damping coefficient; calculating the distance to the nearest obstacle within the preset angle range based on the obstacle position data; calculating the deceleration factor based on the ratio of the nearest obstacle distance to the preset proximity threshold; and calculating the desired motion speed based on the base speed, the distance-corrected speed, and the deceleration factor.

[0022] This invention extracts the resultant force direction as the driving direction, combines the actual and expected distance to solve the distance correction speed, and calculates the base speed, the distance to the nearest obstacle and the deceleration factor in sequence with damping and obstacle parameters. It integrates multiple speed components to obtain the final expected speed, taking into account following correction, gait adaptation and close-range deceleration constraints, thereby improving the driving smoothness and close-range collision avoidance performance of the elderly shopping cart.

[0023] Optionally, the step of calculating the distance correction speed based on the relative position data and the expected following distance includes: extracting the current actual distance from the relative position data; determining the distance deviation based on the current actual distance and the expected following distance; and calculating the distance correction speed based on the distance deviation.

[0024] This invention extracts the real-time actual distance between people and vehicles, compares it with the expected following distance to obtain the distance deviation, and then generates a distance correction speed based on the deviation. It can autonomously adjust the speed by accelerating to approach or decelerating to retreat according to the distance size, thereby improving the dynamic self-calibration capability of the elderly shopping cart following distance.

[0025] Another aspect of the present invention provides an adaptive following control system for a shopping cart based on elderly users, comprising: a processor, an input device, an output device, and a memory, wherein the processor, the input device, the output device, and the memory are interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the adaptive following control method for a shopping cart based on elderly users as described in any of the preceding aspects of the present invention.

[0026] The present invention provides an adaptive following control system for shopping carts for elderly users, which is compact in structure, stable in performance, highly integrated and simple in composition. It can stably execute the adaptive following control method for shopping carts for elderly users provided in the first part of the present invention, further improving the overall applicability and practical application capability of the present invention. Attached Figure Description

[0027] Figure 1 This is a flowchart of an adaptive following control method for shopping carts based on elderly users, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of a shopping cart adaptive following control system based on elderly users, according to an embodiment of the present invention. Detailed Implementation

[0028] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0029] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0030] Please see Figure 1 In order to solve the problems in the prior art, in an alternative embodiment, such as Figure 1 The adaptive following control method for shopping carts based on elderly users, as shown, includes the following steps: Step S1: Obtain the relative position data and relative change rate between the shopping cart and the target elderly user, as well as the obstacle position data and current speed of the shopping cart. The process of obtaining the relative position data and relative rate of change between the shopping cart and the target elderly user, as well as the obstacle position data and current speed of the shopping cart, specifically includes the following sub-steps: Step S101: Obtain the relative position data sequence and relative position historical data of the shopping cart and the target elderly user at a predetermined time interval, as well as the obstacle position data and the current speed of the shopping cart.

[0031] In this embodiment, the shopping cart is equipped with an ultra-wideband base station, a lidar, an ultrasonic sensor, and a motor encoder. The target elderly user wears an ultra-wideband tag bracelet, which has a built-in pedometer that can synchronously output the user's step frequency data. The ultra-wideband tag bracelet communicates wirelessly with the ultra-wideband base station to measure the real-time distance and direction between the shopping cart and the target elderly user. After the shopping cart is started, the target elderly user binds the unique identification code of the ultra-wideband tag bracelet to the current shopping cart by pressing the binding button on the shopping cart handle or selecting a user identifier through the touch screen. This allows the shopping cart to lock onto the target elderly user as the following target, preventing the cart from mistakenly following other users in a multi-user scenario.

[0032] After binding, the shopping cart receives raw distance and azimuth data from the ultra-wideband base station in real time within continuous control cycles (e.g., 50 milliseconds). It also collects relative position data sequences within predetermined time intervals (e.g., a sliding window corresponding to 20 sampling points within the most recent 1.0 second). The window length strikes a balance between suppressing ultra-wideband random ranging noise and maintaining a rapid response to changes in user movement. An excessively long window length introduces tracking lag, while an excessively short length results in insufficient noise suppression. 1.0 second is the optimal value calibrated by engineering. Simultaneously, it acquires historical relative position data within historical time periods. This historical data is used to characterize the previous movement characteristics of the target elderly user. During operation, the motor encoder provides real-time feedback on the shopping cart's current speed, while the lidar and ultrasonic sensors collect real-time obstacle position data around the shopping cart.

[0033] The lidar, as the main sensor, emits a laser beam and receives reflected signals to acquire the distance and orientation information of obstacles within a 360-degree or preset angle range around the shopping cart, forming a polar coordinate point cloud of obstacles. The ultrasonic sensor, as an auxiliary sensor, is installed in front of and on both sides of the shopping cart to supplement the lidar's detection capabilities in near-range blind spots (such as within 0.3 meters directly below the front of the cart) and low-reflectivity objects (such as black shelves and glass cabinet doors). The data from the lidar and ultrasonic sensors are aligned with timestamps and then fused in the shopping cart's coordinate system: the 360-degree detection range is divided into multiple equal-angle sectors. For each sector, if both lidar point cloud and ultrasonic ranging values ​​exist, the obstacle point closer to the shopping cart is retained; if only lidar point cloud or only ultrasonic ranging values ​​exist, the data point is directly retained, forming a unified obstacle point set based on the rule of "sector correspondence, proximity priority".

[0034] Step S102: Determine the maximum allowable change based on the historical relative position data.

[0035] In this embodiment, the maximum permissible change is the maximum allowable value of the relative position difference between two adjacent control cycles, determined using a two-tier mechanism of group initialization and individual adaptation, and always not exceeding a preset physical limit threshold. If the shopping cart does not yet contain historical relative position data for the target elderly user, the preset initial maximum permissible change is used as the effective threshold. The initial maximum permissible change is determined as follows: during the system design phase, a large amount of normal walking gait data of elderly users is collected and statistically analyzed, the distribution characteristics of the relative position change within adjacent control cycles of each user are extracted, and the 95th percentile is taken as the initial value, which is preset as the default startup parameter in the shopping cart system. At the same time, a physical limit threshold is set as a hard upper limit, and the maximum permissible change is taken as the minimum value among the physical limit threshold, the initial maximum permissible change, and the individual adaptive maximum permissible change, ensuring that the effective threshold in any scenario does not exceed the physical limit threshold. The physical limit threshold corresponds to the maximum normal walking speed of the elderly user, calculated with 1.2 m / s corresponding to a 50 ms control cycle, and the maximum displacement is 0.06 m / cycle, used to eliminate noise exceeding the physical limit caused by non-line-of-sight occlusion and multipath reflection. During the operation of the shopping cart, starting from the moment it is linked to the target elderly user, a sliding window continuously records and updates the effective relative position data (i.e., the first data sequence) after amplitude limiting filtering. When the accumulated effective historical data reaches the preset statistical sample size (e.g., 600 sampling points, corresponding to 30 seconds of effective walking time), the 95th percentile of the relative position change in adjacent periods is calculated based on the effective historical data of the target elderly user, generating the individual adaptive maximum allowable change amount, which replaces the initial value as the effective threshold. In subsequent operation, this threshold is continuously updated with the sliding window data, always adapting to the user's current walking rhythm, smoothing out short-term distance fluctuations caused by arm swings and side-stepping to pick up items, while retaining the displacement trend information of normal walking.

[0036] Step S103: Perform amplitude limiting filtering on the relative position data sequence based on the maximum allowable change to obtain the first data sequence.

[0037] In this embodiment, each measured value in the relative position data sequence is compared sequentially with the amplitude-limiting filter output value of the previous control cycle, and the difference between the two is calculated. If the absolute value of the difference is less than or equal to the maximum allowable change, the current measured value is determined to be valid and directly output as the current value of the first data sequence. If the absolute value of the difference is greater than the maximum allowable change, the current measured value is determined to have an abnormal jump due to ultra-wideband signal blockage or multipath effect. At this time, the current output value is replaced with the filter output value of the previous cycle plus the maximum allowable change (with the sign consistent with the direction of the difference), so that the filter output gradually approaches the current measured value with the maximum allowable change as the upper limit, thereby eliminating abnormal jumps while maintaining the continuity of the motion trend. The above comparison and amplitude-limiting operations are sequentially performed on each measured value in the relative position data sequence to form the first data sequence after amplitude-limiting filtering.

[0038] Step S104: Perform mean filtering on the first data sequence to obtain the second data sequence.

[0039] In this embodiment, a fixed-length mean filtering window is set (e.g., 1.0 second, corresponding to 20 sampling points, with a control period of 50ms). A continuous window length of data points, including the current time, is included in the first data sequence within the filtering window. The arithmetic mean of all data points within this window is calculated as the filtered output value for the current time. As new data arrives, the window slides forward, removing the oldest data points and including the newest ones, recalculating the average value within the window. After this mean filtering process, the random noise remaining in the data after amplitude limiting and the high-frequency distance fluctuations caused by arm swinging are effectively smoothed. The average value output by the window is a more stable distance estimate closer to the center of the target elderly user's torso. The above window sliding and average calculation operations are performed sequentially on each data point in the first data sequence to form a second data sequence after mean filtering.

[0040] Step S105: Extract relative position data and relative position data from the previous period from the second data sequence.

[0041] In this embodiment, the data point corresponding to the current control cycle is extracted from the second data sequence after amplitude limiting filtering and mean filtering as the relative position data at the current moment, and the data point corresponding to the previous control cycle is extracted as the relative position data of the previous cycle.

[0042] Step S106: Calculate the relative rate of change based on the relative position data and the relative position data of the previous period.

[0043] In this embodiment, the distance value in the relative position data is subtracted from the distance value in the relative position data of the previous period to calculate the difference. This difference is then divided by the control period to obtain the relative rate of change at the current moment. The relative rate of change characterizes the rate and direction of change of the distance between the shopping cart and the target elderly user over time. When the relative rate of change is negative, it indicates that the distance between the shopping cart and the target elderly user is decreasing (the shopping cart is moving closer to the user); when the relative rate of change is positive, it indicates that the distance between the shopping cart and the target elderly user is increasing (the shopping cart is moving away from the user).

[0044] Step S2: Calculate the resultant force vector based on the relative position data and the obstacle position data.

[0045] The calculation of the resultant force vector based on the relative position data and the obstacle position data specifically includes the following sub-steps: Step S201: Based on the relative position data and the obstacle position data, obtain the user position vector and the obstacle point set.

[0046] In this embodiment, the relative position data is the target polar coordinate information calculated by the ultra-wideband base station, which includes the relative distance and azimuth angle. First, the installation offset and installation deflection angle of the ultra-wideband base station in the shopping cart body coordinate system are obtained. The polar coordinate information is first converted into rectangular coordinates in the ultra-wideband base station's own coordinate system, and then transformed to the shopping cart body coordinate system through rotation and translation to obtain the user position vector with the shopping cart's geometric center as the origin.

[0047] The obstacle location data includes polar coordinate point clouds from a LiDAR radar and ranging values ​​from multiple ultrasonic sensors. The installation offset and deflection angles of the LiDAR and ultrasonic sensors in the shopping cart's coordinate system are obtained: For the LiDAR point cloud, each polar coordinate point is first converted to rectangular coordinates in the LiDAR's own coordinate system, and then uniformly converted to the shopping cart's coordinate system to obtain an initial obstacle point set; for the ultrasonic sensors, based on the preset installation azimuth angle and corresponding ranging value of each probe, they are directly converted to near-range obstacle points in the vehicle's coordinate system. The two types of obstacle points are merged according to the rule of "sector correspondence, proximity priority": obstacle points with smaller ranging values ​​are retained within the corresponding sector, ultimately forming a unified obstacle point set with the shopping cart's geometric center as the origin.

[0048] Step S202: Construct an attraction potential function based on the user position vector.

[0049] The construction of the attraction potential function based on the user's location vector specifically includes the following sub-steps: Step S20201: Obtain the attraction gain base and the ultra-wideband data timeliness status of the relative position data.

[0050] In this embodiment, the attraction gain baseline is a fixed parameter pre-stored in the shopping cart control system's memory. It characterizes the shopping cart's response sensitivity to elderly users. This parameter is calibrated during the system design phase based on the shopping cart's dynamic characteristics and age-appropriate control requirements, and can be directly read from the memory during runtime. The attraction gain baseline is determined using an engineering calibration method: During the system design phase, a reasonable range of attraction gain values ​​is determined through dynamic derivation based on the shopping cart's maximum safe driving speed, maximum following deviation, and equivalent damping coefficient. Then, in a simulated uniform walking test scenario, the gain value is adjusted from small to large, selecting the critical value where the following distance deviation converges rapidly without oscillation as the initial calibration value. Subsequently, specific fine-tuning is performed for typical action scenarios of elderly users, such as slow starts, natural stops, and sideways object retrieval, prioritizing smoothness of movement while ensuring following accuracy, and adapting to the user experience of elderly users. Finally, full-condition verification is completed in typical scenarios such as open aisles, narrow shelf aisles, and pedestrian interference in supermarkets. After confirming no oscillations and no safety risks, the parameter is fixed and stored in the system memory. The timeliness status of ultra-wideband data is maintained in real time by the main control unit based on the timestamp of the last successful reception of the tag signal by the ultra-wideband base station: the current system time is compared with the timestamp. If the time difference is less than the preset timeliness threshold (e.g., 100ms, corresponding to 2 control cycles), the data is determined to be in a fresh state; if the time difference exceeds the timeliness threshold but does not exceed the preset loss threshold (e.g., 1000ms, corresponding to 20 control cycles), the data is determined to be in an outdated state; if the time difference exceeds the loss threshold, the data is determined to be in a lost state.

[0051] In the system design phase, based on the shopping cart's maximum safe driving speed, maximum following deviation, and equivalent damping coefficient, a reasonable range of attraction gain values ​​is determined through dynamic derivation. This includes simplifying the shopping cart's following motion into a first-order damped dynamic model, and establishing a virtual attraction force. For system input, equivalent damping coefficient Characterizing the velocity damping characteristics of the system, shopping cart speed With virtual attraction The relation is satisfied in steady state. Let the preset maximum following deviation be... When the following deviation reaches At that time, the corresponding maximum virtual attraction satisfy ,in To attract gain; Substituting into the steady-state equation, we obtain the corresponding maximum driving speed. ,in Let be the maximum speed at which the shopping cart travels under the maximum following deviation; let the preset maximum safe speed be . To ensure that the actual speed of the shopping cart does not exceed ,make The theoretical upper limit of the attraction gain is derived. The reasonable range for the attraction gain is 50% to 100% of the theoretical upper limit.

[0052] Step S20202: Determine the confidence level based on the timeliness status of the ultra-wideband data.

[0053] In this embodiment, the confidence level is used to quantify the reliability of relative location data, and its value ranges from 0 to 1.0. When the data is fresh, it indicates that the signal communication between the ultra-wideband base station and the tag wristband is normal, and the measurement data is valid in real time. At this time, the confidence level is the maximum value of 1.0. When the data is outdated, it indicates that the signal is temporarily blocked or interfered with, and the positioning data has a certain deviation but still has reference value. At this time, the confidence level is the preset calibration value of 0.6. When the data is lost, it indicates that the signal has been interrupted for a long time or the target has left the effective positioning range. At this time, the confidence level is 0.

[0054] Step S20203: Calculate the attraction gain using the confidence level and the attraction gain base.

[0055] In this embodiment, the attraction gain base is multiplied by the confidence level, and the product is taken as the effective attraction gain for the current control cycle. The attraction gain characterizes the shopping cart's sensitivity to the target elderly user's positional deviation. Its physical meaning is the strength of the attraction generated per unit positional deviation. The larger the attraction gain, the greater the expected acceleration generated by the shopping cart under the same distance deviation, and the faster the following response. Conversely, the smaller the attraction gain, the smoother the shopping cart's following response. By modulating the attraction gain base with the confidence level, adaptive adjustment of control sensitivity based on data quality is achieved.

[0056] Step S20204: Construct an attraction potential function using the user position vector and the attraction gain.

[0057] The attraction potential function satisfies the following formula: , in, To the shopping cart position The attractive potential energy value at that point, To attract gain, Let the current position vector of the shopping cart be _____. Let the location vector be the target elderly user. The Euclidean distance between the shopping cart and the user.

[0058] This formula employs a quadratic potential function design. Its core logic lies in making the attractive potential energy proportional to the square of the distance between the shopping cart and the target elderly user: when the shopping cart moves away from the user, the potential energy increases quadratically, generating a strong pulling force that drives the cart to move quickly closer; as the cart gradually approaches the user, the potential energy decreases smoothly, the pulling force naturally weakens, and the cart approaches the target at a decreasing speed, eventually reaching the point of lowest potential energy equilibrium at the user's location. The entire movement is seamless and oscillating, perfectly aligning with the elderly user's expectation of "rapid response at a distance and smooth approach at a close distance" in a following scenario. Combined with an effective attraction gain dynamically modulated by data confidence, it achieves adaptive smooth following under data quality perception.

[0059] Potential energy is expressed in Newton-meters (N·m). For attraction gain, the unit is Newtons per meter (N / m); and This is a position vector, with units in meters (m). This represents the Euclidean distance, expressed in meters (m). In the formula... The dimensions are Consistent with the dimension of potential energy, for The negative gradient is used to obtain the attraction vector. Its unit is Newton (N).

[0060] Step S203: Construct a repulsive potential function based on the set of obstacle points.

[0061] The construction of the repulsive potential function based on the set of obstacle points specifically includes the following sub-steps: Step S20301: Obtain the equivalent radius of the shopping cart.

[0062] In this embodiment, the equivalent radius of the shopping cart is a fixed parameter pre-stored in the system memory, used to characterize the outline size of the shopping cart on the horizontal plane. Its value is determined by the physical structural size of the shopping cart: half of the larger value between the maximum width and the maximum length of the shopping cart body is taken as the equivalent radius. For example, if the width of the shopping cart body is 0.6m and the length is 0.4m, then the equivalent radius is 0.3m.

[0063] Step S20302: Expand each obstacle point in the obstacle point set radially outward by the equivalent radius to obtain the expanded obstacle point set.

[0064] In this embodiment, for each obstacle point in the obstacle point set, with the current position of the shopping cart as the origin and the direction from the obstacle point to the origin as the radial direction, the obstacle point is moved radially outward by a distance of an equivalent radius to generate the corresponding expanded obstacle point. For the obstacle point located at the... The original position vectors of the obstacle points in the shopping cart's coordinate system The obstacle point, its position after expansion Satisfy the following formula: , in, For the first The position vector of each obstacle point in the shopping cart's coordinate system after inflation. For the first The original position vectors of the obstacle points in the shopping cart's coordinate system. For the shopping cart's equivalent radius, For the first The Euclidean distance from each obstacle point to the origin of the shopping cart's coordinate system. The radial expansion coefficient represents the ratio of the expanded distance to the original distance. The expansion scaling factor is used to stretch the original obstacle points radially outward to their expanded positions. The physical meaning of this expansion operation is to apply the equivalent size of the shopping cart to the obstacles, allowing the subsequently constructed repulsive potential function to use the shopping cart's center of mass as a reference point for obstacle avoidance planning. This is equivalent to treating the shopping cart as a point mass and expanding the obstacle's size outward to solve the obstacle avoidance problem, thus ensuring a safe distance between the actual outline edge of the shopping cart and the obstacles. This expansion operation is performed sequentially on each point in the obstacle point set to obtain the expanded obstacle point set.

[0065] Step S20303: Construct a repulsive potential function based on the expanded obstacle point set.

[0066] The construction of the repulsive potential function based on the expanded obstacle point set specifically includes the following sub-steps: Step S2030301: Obtain the repulsion gain and spatial influence scale parameters.

[0067] In this embodiment, the repulsion gain and the spatial influence scale parameter are both fixed parameters pre-stored in the memory of the shopping cart control system: the repulsion gain is used to control the shopping cart's obstacle avoidance response intensity. The greater the gain, the stronger the repulsion force generated by the obstacle at the same distance, and the greater the obstacle avoidance action amplitude; the spatial influence scale parameter is used to control the spatial range of the repulsion potential field. The greater the parameter, the farther the effective influence distance of the repulsion force, and the earlier the shopping cart triggers the avoidance action.

[0068] Two parameters were determined through engineering calibration during the system design phase. Prioritizing aging-friendly smoothness, and ensuring a safe margin while pursuing smooth movement, a single-variable control method was used for step-by-step calibration: First, a simulated supermarket test environment was built, setting a 1.2m wide standard shelf aisle as a typical narrow aisle scenario. The repulsion gain was fixed as an empirical initial value (e.g., a baseline value calibrated using population statistical data). The spatial influence scale parameter was gradually adjusted from small to large, and the shopping cart's passage and obstacle avoidance performance were observed. The critical value that "the shopping cart can detect obstacles in advance and smoothly navigate around them, without causing passage jams due to the superposition of repulsion forces on both sides of the narrow aisle caused by an excessively large range of action" was selected as the spatial influence. The calibration values ​​of the scale parameters are then fixed. Subsequently, the repulsion gain is gradually adjusted from small to large while simultaneously detecting the minimum distance between the vehicle edge and obstacles. The gain value that "meets the preset safety margin (e.g., not less than 0.1m), the bypass trajectory is smooth without oscillation, and there are no obvious lateral abrupt changes" is selected as the calibration value of the repulsion gain. Finally, full-condition verification is completed in multiple typical supermarket scenarios such as open main aisles, narrow shelf aisles, corner obstacles, and low obstacles. After confirming that the obstacle avoidance can be safe and smooth in all scenarios, the two parameters are solidified as fixed values ​​and stored in the system memory. They can be directly read from the memory during runtime.

[0069] Step S2030302: Calculate the obstacle point repulsion potential function components based on the repulsion gain, the spatial influence scale parameter, and the expanded obstacle point set.

[0070] In this embodiment, for each obstacle point in the expanded obstacle point set, the repulsive potential function component corresponding to that point is calculated. The repulsive potential function component of the obstacle point satisfies the following formula: , in, For the first An inflated obstacle point is located in the shopping cart. The repulsive potential component generated at that location, To exclude gain, Let the position of the shopping cart in a two-dimensional plane be the variable. For the first The position vectors of the inflated obstacle points For shopping cart and the first The Euclidean distance between the expanded obstacle points This is the spatial scale parameter. The Gaussian potential function reaches its maximum value at the obstacle point. The potential energy decreases exponentially with increasing distance; the closer the distance, the higher the potential energy and the stronger the repulsive force felt by the shopping cart. Conversely, the potential energy approaches zero and the repulsive force becomes negligible with increasing distance. Furthermore, the function is continuously differentiable at any position, ensuring the shopping cart will not experience abrupt forces at obstacle edges, thus guaranteeing smooth obstacle avoidance. The above calculation is performed on each obstacle point in the expanded obstacle point set to obtain the repulsive potential function components corresponding to each obstacle point.

[0071] For the first The repulsive potential component corresponding to each obstacle point is expressed in Newton-meter (N·m). The repulsive gain is expressed in Newton-meters (N·m). and This is a position vector, with units in meters (m). This is the Euclidean distance, in meters (m). This is a parameter representing the spatial scale of influence, measured in meters (m). (Exponential term) The numerator and denominator are both dimensionless. After division, the result is a dimensionless quantity, therefore and Both have the same dimensions, N·m. Regarding the total repulsive potential function... Finding the negative gradient yields the repulsive force vector. The unit is Newton (N). Similar to the aforementioned attractive potential function, the unit of the repulsive potential function is also N·m, and the unit of the repulsive force vector obtained after calculating the negative gradient is also N.

[0072] Step S2030303: The repulsion potential function components of the obstacle point are superimposed to obtain the repulsion potential function.

[0073] In this embodiment, the repulsion potential function components of all obstacle points calculated in step S2030302 are superimposed by scalar, that is, the repulsion potential function components of all obstacle points are summed to obtain the total repulsion potential function: , in, Place all obstacle points in the shopping cart location. The total repulsive potential energy generated at that point, This represents the total number of obstacle points in the expanded obstacle point cluster. For the first An inflated obstacle point is located in the shopping cart. The repulsive potential component generated at that location, To exclude gain, For the first The position vectors of the inflated obstacle points For shopping cart and the first The Euclidean distance between the expanded obstacle points The spatial scale parameter represents the total repulsive potential function. This function combines the repulsive effects of all obstacles into a unified scalar field. The repulsive force experienced by the shopping cart in this potential field is the combined effect of the repulsive forces of each obstacle. The repulsive forces in multiple directions are superimposed to form the resultant force, pointing towards an open area with a low density of obstacles.

[0074] Step S204: Calculate the negative gradient of the attraction potential function to obtain the attraction vector.

[0075] In this embodiment, the attraction potential function The negative gradient yields an attraction vector that satisfies the following formula: , in, As an attraction vector, For gradient operators, To attract gain, Let the current position vector of the shopping cart be _____. Let the location vector be the target elderly user. Let this be the vector pointing from the shopping cart to the target elderly user. In the shopping cart's coordinate system, the current position of the shopping cart is the origin. The attraction vector simplifies to Its direction is from the shopping cart to the target elderly user, and its size is proportional to the distance between the shopping cart and the user. The farther the distance, the greater the attraction, and the closer the distance, the less the attraction. When the shopping cart reaches the user's position, the attraction is zero.

[0076] Step S205: Calculate the negative gradient of the repulsive potential function to obtain the repulsive force vector.

[0077] In this embodiment, the repulsive potential function The negative gradient yields a repulsive force vector that satisfies the following formula: , In the coordinate system of the shopping cart itself, the current position of the shopping cart is the origin. The repulsive force vector simplifies to: , in, It is the repulsive force vector. For gradient operators, To exclude gain, For spatial influence scale parameters, Let the current position vector of the shopping cart be _____. For the first The position vectors of the inflated obstacle points For from the first The vector pointing from the inflated obstacle point to the shopping cart. For shopping cart and the first The Euclidean distance between the expanded obstacle points This represents the total number of obstacle points in the expanded obstacle point set.

[0078] Its direction is the result of the vector superposition of the repulsion directions of each obstacle point, and the whole direction points to the open area with low obstacle distribution density; the closer the obstacle is, the larger the exponent term is and the stronger the repulsion force is; the farther the obstacle is, the exponent term approaches zero and the repulsion force can be ignored.

[0079] Step S206: The attractive force vector and the repulsive force vector are vector-added to obtain the resultant force vector.

[0080] The resultant force vector satisfies the following formula: , in, The resultant force vector, As an attraction vector, This represents the repulsive force vector. In the shopping cart's coordinate system, the attractive force vector points towards the target elderly user, while the repulsive force vector points towards an open area with low obstacle density. The direction of the resultant force vector, when superimposed, represents the optimal overall direction the shopping cart should take at the current moment (approaching the user while avoiding obstacles). The magnitude of the resultant force vector represents the degree of deviation from the equilibrium state; a larger magnitude indicates a greater deviation, requiring a faster response from the shopping cart. This vector addition operation achieves a natural fusion of following the target and avoiding obstacles, without the need for mode switching conditions or priority arbitration. The shopping cart continuously and smoothly adjusts its direction of movement based on the direction of the resultant force vector.

[0081] Step S3: Calculate the current user speed of the target elderly user based on the relative rate of change and the current speed of the shopping cart.

[0082] In this embodiment, the component of the current shopping cart speed along the shopping cart-user connection direction is first calculated based on the current shopping cart speed. Then, the current user speed is calculated based on the component of the current speed along the shopping cart-user connection direction and the relative rate of change.

[0083] Current speed of users targeting elderly users Satisfy the following formula: , in, The current speed is designed for target elderly users. The component of the shopping cart's current speed along the direction of the line connecting the shopping cart and the user. This is the relative rate of change (i.e., the rate of change of the distance between the shopping cart and the user; positive for increasing distance and negative for decreasing distance). When the relative rate of change... A negative value indicates that the distance between the shopping cart and the user is decreasing, meaning the shopping cart is approaching the user, and the shopping cart's speed is greater than the user's current speed; when... When the value is positive, it indicates that the distance between the shopping cart and the user is increasing, meaning the user is moving away from the shopping cart; at this time, the user's current speed is greater than the shopping cart's current speed. When the value is zero, it indicates that the distance between the shopping cart and the user remains constant, and the user's current speed is equal to the shopping cart's current speed. In the supermarket following scenario, the walking direction of the target elderly user is basically consistent with the direction of the line connecting the shopping cart and the user. Therefore, the above radial velocity component can approximate the user's actual walking speed, and the calculation result is used to determine the expected following distance in subsequent steps.

[0084] Step S4: Determine the desired following distance based on the user's current speed, and determine the desired motion direction and desired motion speed based on the resultant force vector and the desired following distance.

[0085] Determining the desired following distance based on the user's current speed specifically includes the following sub-steps: Step S401: Obtain the minimum following distance and reaction time coefficient.

[0086] In this embodiment, both the minimum following distance and the reaction time coefficient are fixed parameters pre-stored in the shopping cart control system's memory, which can be directly read from the memory during runtime. The minimum following distance is the minimum safe distance allowed between the shopping cart and the target elderly user. Its value is determined based on the elderly user's physical activity space during normal shopping, with a typical value of 0.5 meters. This ensures that the shopping cart will not cause collisions or a feeling of oppression when the elderly user stops, turns around, or reaches for items, while also not affecting the user's convenience due to excessive distance. The reaction time coefficient characterizes the shopping cart's response lead to changes in the elderly user's speed. Its value is determined based on the elderly person's reaction time characteristics during walking, with a typical value of 1 second. This is used to achieve a linear adjustment of the desired following distance with the user's speed: a larger reaction time coefficient results in a larger following distance when the user walks quickly, leading to higher safety; a smaller reaction time coefficient results in a smaller following distance when the user walks slowly, leading to better following tightness.

[0087] Step S402: Calculate the expected following distance based on the user's current speed, the minimum following distance, and the reaction time coefficient.

[0088] In this embodiment, the desired following distance is satisfied by the following formula: , in, To determine the desired distance, To minimize the following distance, This is the reaction time coefficient. This represents the user's current speed.

[0089] The expected following distance consists of two parts, the first part Based on the basic safety distance, ensuring that the shopping cart does not cause collisions or a feeling of oppression for elderly users when they stop or turn around, Part Two The following distance is dynamically adjusted and proportional to the user's current speed. The faster the user walks, the greater the following distance the shopping cart maintains, providing the user with more time to react and a safe buffer. The slower the user walks, the smaller the following distance the shopping cart maintains, preventing the following from failing due to excessive distance.

[0090] Determining the desired motion direction and desired motion speed based on the resultant force vector and the desired following distance specifically includes the following sub-steps: Step S411: Extract the vector direction from the resultant force vector to obtain the desired motion direction.

[0091] In this embodiment, the resultant force vector It is a two-dimensional vector containing direction and magnitude information. Its direction is determined by calculating the angle between the vector and the positive X-axis of the shopping cart's coordinate system (i.e., the direction the cart is facing). ,in Let the desired direction angle of motion be... and These are the components of the resultant force vector along the X and Y axes of the shopping cart's coordinate system, respectively. This is the arctangent function in the four quadrants, and its return value range is... Negative values ​​indicate a right turn, and positive values ​​indicate a left turn. When the magnitude of the resultant force vector is not zero, the included angle is the desired direction of movement for the shopping cart; when the magnitude of the resultant force vector is zero, it indicates that the current following and obstacle avoidance forces are balanced, there is no clear turning requirement, the desired direction of movement follows the direction of the previous moment, and the movement speed is determined by the subsequent speed calculation stage.

[0092] Step S412: Calculate the distance correction speed based on the relative position data and the expected following distance.

[0093] The calculation of the distance correction speed based on the relative position data and the expected following distance specifically includes the following sub-steps: Step S41201: Extract the current actual distance from the relative position data.

[0094] In this embodiment, the relative position data includes the real-time distance and azimuth information between the shopping cart and the target elderly user. The distance component is directly extracted from the relative position data at the current moment as the current actual distance between the shopping cart and the target elderly user, denoted as . .

[0095] Step S41202: Determine the distance deviation based on the current actual distance and the expected following distance.

[0096] In this embodiment, the distance deviation satisfies the following formula: , in, For distance deviation, This is the current actual distance. To determine the desired following distance. When When the actual distance is greater than the expected following distance, it means the shopping cart is too far from the user and needs to be moved closer; when When the actual distance is less than the expected following distance, it means the shopping cart is too close to the user and needs to slow down or stop; when When the current actual distance is equal to the expected following distance, the shopping cart and the user are at the ideal following distance.

[0097] Step S41203: Calculate the distance correction speed based on the distance deviation.

[0098] In this embodiment, the distance correction speed satisfies the following formula: , in, For distance-corrected speed, This is the distance deviation ratio coefficient. This represents the distance deviation. When When the shopping cart is too far from the user, the distance correction speed is positive, generating a correction amount to accelerate the movement closer; when When the shopping cart is too close to the user, the distance correction speed is negative, resulting in a deceleration or stop. At this point, the distance correction speed is zero, and there is no need to correct the base speed.

[0099] Distance Deviation Ratio Coefficient The value of determines the sensitivity of the distance correction speed to distance deviation, and it is determined as follows: In the system design phase, it is first derived through the shopping cart dynamics model and maximum acceleration capability. Theoretical upper limit: distance correction speed Taking the derivative with respect to time, the corrected acceleration corresponding to the distance-corrected velocity is: ,in The rate of change of distance deviation; to ensure that the correction acceleration does not exceed the maximum acceleration capacity of the shopping cart. , must meet ,in The maximum absolute value of the rate of change of distance deviation can be conservatively estimated using the user's maximum walking speed, for example, taking... , ,but To ensure the shopping cart doesn't oscillate due to overcorrection at this gain, a standard test sequence was then simulated in a simulation environment, including user walking at a constant speed, accelerating, decelerating, and stopping, to ensure the distance deviation converges to the baseline expected following distance (i.e., the typical expected following distance value preset during the system design phase, such as 0.7m) within 2 seconds. Within the specified range, and with no overshoot or oscillation in the speed response as the criterion, the minimum gain value meeting these conditions was selected as the calibration value. Finally, verification was completed under multiple different speed scenarios to ensure that the speed is within the user's specified range. to hour, All maintained stable convergence performance. This was determined through the above calibration procedure. The range of values ​​is to It can ensure that the amplitude of the distance correction speed does not exceed 50% of the base speed, thus achieving smooth correction of distance deviation and avoiding motion oscillation caused by over-correction.

[0100] Step S413: Obtain the equivalent damping coefficient, preset angle range, and preset near distance threshold.

[0101] In this embodiment, the equivalent damping coefficient The equivalent damping coefficient is used to map the resultant force modulus to the base velocity, and its value determines the shopping cart's response sensitivity to the virtual resultant force. The equivalent damping coefficient is obtained as follows: first, the nominal equivalent damping coefficient is read from the system memory. The system first sets a preset angle range and a preset proximity threshold; then it determines whether the adaptive modulation triggering conditions are met. When the target elderly user simultaneously meets the following conditions, it is determined that the user is in a stable walking state and enters the adaptive modulation process: Current user speed To eliminate the state of being stopped; The standard deviation of the relative distance data within the last 2 seconds is less than a preset threshold to exclude non-linear walking states such as turning around or turning to the side to pick up objects. The stride estimation value is continuously valid. The ultra-wideband wristband pedometer detects at least one valid step frequency signal in the most recent second (corresponding to the gait characteristics of the lowest step frequency of about 1Hz for elderly users, ensuring that the gait characteristics have basic statistical significance) to exclude signal loss.

[0102] If any of the above conditions are not met, the user is determined not to be in a stable walking state, and the nominal value is used directly. No modulation is triggered.

[0103] When adaptive modulation is triggered, the current step frequency is obtained from the pedometer built into the ultra-wideband wristband. Utilize the user's current speed Calculate the current stride The average stride length within the most recent 3-second sliding window is used as the baseline stride length. (A 3-second window covers approximately 3-6 steps, resulting in better statistical stability), calculate the stride deviation rate. The equivalent damping coefficient satisfies the following formula: , And the modulated damping coefficient is limited to to Within the specified range, the equivalent damping coefficient satisfies the following formula: , in, The effective equivalent damping coefficient. The nominal equivalent damping coefficient is... For preset modulation intensity coefficient, This represents the stride deviation rate; a positive value indicates a decrease in stride length and user deceleration, while a negative value indicates an increase in stride length and user acceleration. When... When this occurs, it is considered a normal gait fluctuation, does not trigger modulation, and the damping coefficient remains at its nominal value. ;when and When the value is positive, it indicates that the user is decelerating, the damping coefficient increases, the base speed decreases, and the shopping cart response becomes smoother; when... and A negative value indicates that the user is accelerating, the damping coefficient decreases, the base speed increases, and the shopping cart response becomes more sensitive. (Clamping operation) Ensure that the damping coefficient after modulation does not exceed the physically reasonable range to avoid system oscillation caused by over-modulation.

[0104] The calibration method for the preset modulation intensity coefficient α is as follows: During the system design phase, representative elderly subjects are selected in a simulated test environment, and data is collected under a standard test sequence including constant speed walking, accelerated walking, and deceleration and stopping. The nominal damping coefficient is fixed as the initial design value, and α is gradually increased from 0.2, increasing by 0.1 each round. The speed response curve of the shopping cart is observed: When α is too small, the damping coefficient lags significantly in tracking stride changes. When the user accelerates, the shopping cart responds slowly and the distance increases. When the user decelerates, the shopping cart brakes untimely and the feeling of approach is obvious. When α gradually increases to around 0.5~0.6, the phase difference between the speed response curve and the user speed change curve is the smallest, and there is no overshoot or oscillation. When α is further increased to above 0.8, the damping coefficient is too sensitive to small fluctuations in stride, and the shopping cart speed fluctuates significantly and adjusts back and forth when the user is walking at a constant speed. Therefore, the calibration value of α is determined to be the critical value that minimizes the speed response delay under the premise of no overshoot and no oscillation. The typical calibration result is 0.5~0.6. Different models can be finely adjusted within this range according to the shopping cart weight and motor response characteristics. After being solidified into system parameters, they are stored in memory and do not need to be adjusted online during operation.

[0105] The modulated damping coefficient is limited to to Within this range, the reason is as follows: This range is determined based on the linear operating range of the shopping cart dynamics—the shopping cart motor drive system within the nominal damping coefficient. Within a range of ±40%, the response remains linear. Beyond this range, the motor enters the saturation or dead zone, and the control linearity decreases significantly. This range also covers the rate of change of inertia between empty and fully loaded shopping carts (typically not exceeding ±30%), ensuring that damping adjustment matches the range of changes in the physical characteristics of the shopping cart itself. From an aging-friendly safety perspective, the upper limit... Ensure the shopping cart retains a minimum level of responsiveness under all circumstances, avoiding situations where excessive damping leads to insufficient base speed and tracking failure. To ensure the shopping cart is not overly sensitive under any circumstances, and to prevent gait fluctuations from being amplified into speed oscillations due to insufficient damping; the specific value of this range was verified and confirmed through extreme condition testing during the system design phase—gradually increasing the damping coefficient until the response delay exceeds 0.5 seconds (the critical value is approximately...). After taking a safety margin, it is determined to be Gradually reduce the damping coefficient until speed fluctuations occur during constant-speed walking (critical value approximately...). After taking a safety margin, it is determined to be This ensures that the shopping cart maintains stable and smooth following motion under all normal operating conditions.

[0106] Equivalent damping coefficient The value of directly affects the base speed, and thus the final desired movement speed. This is dynamically adjusted through the aforementioned gait adaptive modulation mechanism. This achieves adaptive matching of the base speed to the user's walking rhythm. When the user's stride length decreases (deceleration or preparation to stop), Increase The shopping cart slows down gradually in advance; when the user's stride length increases (accelerates from a standstill), Decrease Increase the size of the shopping cart so it can respond quickly and avoid increasing the distance between carts.

[0107] The preset angle range is used to define the angle of the fan-shaped detection area in front of the shopping cart. Obstacle points are filtered within this area to calculate the distance to the nearest obstacle. A typical value is ±30° to ±45° in front, with the specific value determined based on the shopping cart's motion characteristics and the typical distribution of obstacles in a supermarket setting. An excessively large angle range may introduce interference from irrelevant obstacles on the sides, while an excessively small range may miss dangerous obstacles directly in front. The preset proximity threshold is used to determine whether an obstacle has entered the shopping cart's proximity deceleration zone. Deceleration is triggered when the nearest obstacle distance is less than this threshold. A typical value is 0.5m to 0.8m, with the specific value determined based on the shopping cart's braking performance and the safety needs of elderly users. All three parameters are calibrated during the system design phase based on the shopping cart's dynamic characteristics and age-appropriate control requirements, and then stored as default values ​​in the system memory. The equivalent damping coefficient is adaptively modulated in real-time based on the user's gait characteristics during operation, while the other parameters use their default values ​​directly.

[0108] Step S414: Calculate the resultant force modulus based on the resultant force vector, and calculate the base velocity based on the resultant force modulus and the equivalent damping coefficient.

[0109] In this embodiment, the resultant force vector Two-dimensional vector Its resultant force modulus Satisfy the following formula: , in and The resultant force vectors in the shopping cart's body coordinate system are respectively shaft and Components on the axis, The resultant force magnitude represents the magnitude of the current resultant force vector. A larger magnitude indicates a more severe deviation of the shopping cart from its equilibrium state, requiring a faster response. The resultant force magnitude... Divide by the equivalent damping coefficient The resulting base velocity satisfies the following formula: , in, The base velocity is a scalar value representing the basic response velocity of the shopping cart under the current magnitude of the resultant force and damping coefficient. When the equivalent damping coefficient... When the resultant force modulus is fixed, the larger the resultant force modulus, the greater the base velocity. When the resultant force modulus is fixed, the larger the equivalent damping coefficient, the smaller the base velocity, and the smoother the shopping cart response. The smaller the equivalent damping coefficient, the greater the base velocity, and the more sensitive the shopping cart response.

[0110] Step S415: Calculate the distance to the nearest obstacle within the preset angle range based on the obstacle position data.

[0111] In this embodiment, an obstacle point set is calculated based on obstacle location data. Each point in the obstacle point set contains distance and azimuth information (relative to the shopping cart's coordinate system). A preset angle range is used to define the angle of the fan-shaped detection area in front of the shopping cart. Each obstacle point in the obstacle point set is traversed, and it is determined whether the azimuth of each point is within the preset angle range (i.e., the fan-shaped area covered by half of the preset angle range to the left and right of the shopping cart's facing direction). Only when the azimuth of an obstacle point is within this preset angle range is the obstacle point included in the candidate set. Among all candidate obstacle points, the obstacle point closest to the shopping cart is selected, and its distance value is the nearest obstacle distance within the preset angle range, denoted as . If no obstacle is detected within the preset angle range, the distance to the nearest obstacle is set to the preset maximum value (e.g., ...). (), indicating that there are no effective obstacles in the sector area ahead.

[0112] Step S416: Calculate the deceleration factor based on the ratio of the distance to the nearest obstacle to the preset proximity threshold.

[0113] In this embodiment, the deceleration factor satisfies the following formula: , in, This is a deceleration factor, with a value ranging from 0 to 1. The distance to the nearest obstacle. This is a preset proximity threshold. When... When the ratio is 1 or greater than 1, take the minimum value. This indicates that the obstacle ahead is far away and there is no need to slow down; when When the ratio is less than 1, This indicates that an obstacle ahead has entered the close-range deceleration zone. The closer the obstacle, the smaller the deceleration factor, and the more noticeably the shopping cart should slow down. When it approaches 0, The deceleration factor approaches 0, indicating the shopping cart is nearing a stop. This deceleration factor actively reduces the cart's speed when an obstacle is in front, with the deceleration increasing smoothly and continuously as the obstacle approaches, preventing sudden braking. If no obstacle is detected within the preset angle range ahead, Take the preset maximum value (e.g., 3.0m). , It does not produce a speed reduction effect.

[0114] The preset proximity threshold is determined during the system design phase based on the shopping cart's maximum braking deceleration. (Determined by motor performance and ground adhesion coefficient) and system delay (Including sensor sampling period, communication delay, and control period), calculate the theoretical safe distance. The theoretical safe distance satisfies the following formula: , Calculate the theoretical safe distance, where, The first term of the formula represents the maximum speed at which the shopping cart can run. The distance the shopping cart travels during the corresponding system delay period, the second item This corresponds to the braking distance required for the shopping cart to come to a complete stop from its maximum speed. Based on the calculated theoretical value, verification and fine-tuning were conducted in a simulated supermarket test environment, prioritizing smoothness and aging adaptability. The maximum operating speed of the shopping cart was fixed, and the proximity threshold was gradually increased starting from 80% of the theoretical value. A critical value was selected as the calibration value for the preset proximity threshold, which is "the shopping cart can brake smoothly in advance when encountering a static obstacle in front, without obvious sudden braking sensation, and will not frequently trigger unnecessary deceleration during normal following due to an excessively high threshold." The typical range of values ​​is 0.5m to 0.8m.

[0115] Step S417: Calculate the desired motion speed based on the base speed, the distance correction speed, and the deceleration factor.

[0116] In this embodiment, the desired motion speed satisfies the following formula: , in, For the desired speed of motion, As a deceleration factor, Based on speed, This is the speed correction for distance.

[0117] The base speed and distance correction speed are first added together. The physical meaning of their sum is: the base speed determines the degree to which the shopping cart deviates from its equilibrium state in response to its current position (the greater the net force, the faster it moves); the distance correction speed allows the shopping cart to actively adjust its distance from the user (accelerating to move closer when too far away, and decelerating or stopping when too close). Together, these two determine the shopping cart's baseline speed when there are no obstacles. The sum is then multiplied by a deceleration factor. It enables active deceleration when there are obstacles nearby, and deceleration when there are no obstacles or the obstacles are far away. The speed is unaffected when the obstacle enters the close deceleration zone. The speed decreases proportionally, with the speed decreasing further as the obstacle approaches, achieving smooth deceleration and obstacle avoidance. The calculated desired speed... This is a scalar value, representing the magnitude of the shopping cart's velocity in the desired direction of movement.

[0118] Step S5: Determine the shopping cart speed control command based on the desired direction of movement and the desired speed of movement.

[0119] In this embodiment, firstly, a desired velocity vector is constructed using the desired direction of motion as the velocity direction and the desired velocity as the velocity amplitude: , in, Let the desired two-dimensional velocity vector be defined in the coordinate system of the shopping cart itself. Let the desired motion velocity scalar amplitude be... Let the desired direction angle of motion be... and These are the cosine and sine values ​​of the direction angle, respectively, used to decompose the scalar velocity into... Axle (direction of vehicle front) and The velocity component along the axis (lateral direction).

[0120] Subsequently, each component of the desired velocity vector is subjected to a first-order low-pass filter to eliminate high-frequency jitter in the velocity command, making the shopping cart movement transition smoother. , in, This is the velocity vector after filtering in the current period. These are the filter coefficients (ranging from 0.2 to 0.5, determined by system calibration). Let V be the expected velocity vector for the current cycle. This is the velocity vector after filtering in the previous cycle.

[0121] Next, the filtered velocity vector is subjected to acceleration limiting. , Each component is constrained to determine the maximum change in a single period: , in, The velocity vector after amplitude limiting. This is the velocity vector after filtering in the previous cycle. This is the velocity vector after filtering in the current period. To be Limited to Clamping function within the interval, The preset maximum acceleration (typical value 0.3~0.5 m / s²) is used. 2 (Aging-friendly calibration) To control the cycle, The maximum allowable speed change within a single cycle ensures a smooth change in shopping cart speed, avoiding sudden acceleration or deceleration that might startle elderly users.

[0122] Next, apply a backtracking safety constraint: Based on the user's position vector, determine whether the target elderly user is located within a ±90° sector area in front of the shopping cart and whether the current actual distance is less than a preset stop threshold (typically 0.5m). If so, set the backward component along the user's direction in the limited velocity vector to zero. , in, The velocity vector after safety constraints, The velocity vector after amplitude limiting. Let be the unit vector pointing from the shopping cart to the user. This is the projection of the velocity vector in the user's direction (a negative value indicates the presence of a backward component). The magnitude of the backward component is used to ensure that the shopping cart will not collide with the user when the user is close in front.

[0123] Finally, the speed command after safety constraints is mapped to the motor control signals for the two-wheel differential drive. This is done using the speed vector... The axle component is used as the desired linear velocity of the vehicle body, and the desired angular velocity is calculated using proportional control, with the desired motion direction angle as the heading deviation. , in, The desired linear velocity of the shopping cart (in the direction of travel of the cart). For the velocity vector after safety constraints Axial components; For the expected angular velocity of the shopping cart, This is the heading proportional gain (determined by system calibration; the larger the value, the faster the steering response). The desired direction angle of movement (positive value corresponds to left turn, negative value corresponds to right turn).

[0124] Based on the kinematic model of a two-wheel differential chassis, linear velocity and angular velocity are converted into the desired velocities of the left and right wheels: , in, Let the desired linear velocity of the revolver be... Let the desired linear velocity of the right wheel be... The wheel track of the shopping cart (the distance between the center of the left and right wheels, a physical parameter of the shopping cart) is used to further convert the speed of the left and right wheels into PWM duty cycle signals for the motor and output them to the drive module as the final speed control command for the shopping cart.

[0125] like Figure 2 As shown, in another aspect, the present invention also provides an adaptive following control system for a shopping cart based on elderly users, comprising: a processor, an input device, an output device, and a memory, wherein the processor, the input device, the output device, and the memory are interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to call the program instructions to execute relevant steps in a relevant embodiment of the adaptive following control method for a shopping cart based on elderly users of the present invention.

[0126] This invention provides an adaptive following control system for shopping carts designed for elderly users. The functional components can be integrated into a single processing unit, or each component can exist independently, or two or more components can be integrated into one unit. The integrated components can be implemented in hardware or software. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this invention and are not intended to limit it. Although the invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this invention, and all should be covered within the scope of the claims and specification of this invention.

Claims

1. A shopping cart adaptive following control method for elderly users, characterized in that, The method includes: Acquire the relative position data and relative rate of change between the shopping cart and the target elderly user, as well as the obstacle position data and current speed of the shopping cart; Calculate the resultant force vector based on the relative position data and the obstacle position data; Calculate the target elderly user's current speed based on the relative rate of change and the current speed of the shopping cart; The desired following distance is determined based on the user's current speed, and the desired direction and speed of motion are determined based on the resultant force vector and the desired following distance. The shopping cart speed control command is determined based on the desired direction of movement and the desired speed of movement.

2. The adaptive following control method for shopping carts based on elderly users according to claim 1, characterized in that, The acquisition of the relative position data and relative rate of change between the shopping cart and the target elderly user, as well as the obstacle position data and current speed of the shopping cart, includes: Acquire a sequence of relative position data between the shopping cart and the target elderly user at predetermined time intervals, historical relative position data, obstacle location data of the shopping cart, and the current speed of the shopping cart; The maximum permissible change is determined based on the historical data of the relative position; The first data sequence is obtained by performing amplitude limiting filtering on the relative position data sequence based on the maximum allowable change. The second data sequence is obtained by performing mean filtering on the first data sequence; Extract the relative position data and the relative position data of the previous period from the second data sequence; The relative rate of change is calculated based on the relative position data and the relative position data of the previous period.

3. The adaptive following control method for a shopping cart based on elderly users according to claim 1, characterized in that, The calculation of the resultant force vector based on the relative position data and the obstacle position data includes: Based on the relative position data and the obstacle position data, the user position vector and the obstacle point set are obtained; Constructing an attraction potential function based on the user location vector includes: Obtain the ultrawideband data timeliness status of the attraction gain base and the relative position data; The confidence level is determined based on the timeliness status of the ultra-wideband data; The attraction gain is calculated using the confidence level and the attraction gain base. An attraction potential function is constructed using the user location vector and the attraction gain; Constructing a repulsive potential function based on the set of obstacle points includes: Get the equivalent radius of the shopping cart; Expand each obstacle point in the obstacle point set radially outward by the equivalent radius to obtain the expanded obstacle point set; Constructing a repulsive potential function based on the expanded obstacle point set includes: Obtain the repulsion gain and spatial influence scale parameters; The components of the obstacle point repulsion potential function are calculated based on the repulsion gain, the spatial influence scale parameter, and the expanded obstacle point set. The repulsive potential function is obtained by superimposing the components of the repulsive potential function at the obstacle point. Calculate the negative gradient of the attraction potential function to obtain the attraction vector; Calculate the negative gradient of the repulsive potential function to obtain the repulsive force vector; The resultant force vector is obtained by vector addition of the attraction force vector and the repulsion force vector.

4. The adaptive following control method for a shopping cart based on elderly users according to claim 1, characterized in that, Determining the expected following distance based on the user's current speed includes: Obtain the minimum following distance and reaction time coefficient; The expected following distance is calculated based on the user's current speed, the minimum following distance, and the reaction time coefficient.

5. The adaptive following control method for shopping carts based on elderly users according to claim 1, characterized in that, The step of determining the desired motion direction and desired motion speed based on the resultant force vector and the desired following distance includes: The desired motion direction is obtained by extracting the vector direction from the resultant force vector. Calculate the distance correction speed based on the relative position data and the expected following distance; Obtain the equivalent damping coefficient, preset angle range, and preset proximity threshold; The resultant force modulus is calculated based on the resultant force vector, and the foundation velocity is calculated based on the resultant force modulus and the equivalent damping coefficient. Calculate the distance to the nearest obstacle within the preset angle range based on the obstacle location data; The deceleration factor is calculated based on the ratio of the distance to the nearest obstacle to the preset proximity threshold. The desired speed is calculated based on the base speed, the distance-corrected speed, and the deceleration factor.

6. The adaptive following control method for a shopping cart based on elderly users according to claim 5, characterized in that, The calculation of distance correction speed based on the relative position data and the expected following distance includes: Extract the current actual distance from the relative position data; The distance deviation is determined based on the current actual distance and the expected following distance; The distance correction speed is calculated based on the distance deviation.

7. A shopping cart adaptive following control system for elderly users, characterized in that, include: The system includes a processor, an input device, an output device, and a memory, all interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute a shopping cart adaptive following control method for elderly users as described in any one of claims 1 to 6.