A control method for rubber tracked indoor engineering vehicles

Abstract

The present invention provides a method for controlling rubber tracks on indoor engineering vehicles. By acquiring data from multimodal sensors and track deviation data, the method accurately identifies the ground material and adaptively adjusts the contact area and tension of the rubber track to effectively compensate for track deviation. The present invention is able to sense changes in ground material in real time and dynamically adjust the contact area and tension of the rubber track accordingly. This technology not only enhances the vehicle's adaptability on different ground materials, ensuring that indoor engineering vehicles follow a preset walking track, but also reduces slippage and collision risks, thereby improving the overall operating efficiency and safety of indoor engineering vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering vehicles, and in particular to a method for controlling a rubber track on an indoor engineering vehicle. Background Art

[0002] In the existing technology, Chinese invention patent CN103754281B provides a track wheel with adjustable contact area and tension. In addition to ensuring the original steering, braking and transmission functions, by controlling the filling and discharge of oil in the double-acting hydraulic cylinder in the contact area adjustment mechanism, it can achieve adjustment of different contact areas under three road conditions: highway, off-road and extreme conditions; by controlling the tensioning cylinder in the track tensioning mechanism, the rubber track can be tightened and loosened, ensuring that the track will not fall off during driving or turning, thereby improving the vehicle's ability to adapt to road conditions.

[0003] Rubber tracks are more suitable than other types of tracks (such as steel tracks) in certain scenarios. For example, food processing plants use rubber tracks for their engineering vehicles, which are easy to clean, meet hygiene standards, and ensure food safety. Wood processing plants and paper mills also use rubber tracks for their engineering vehicles, as they are non-sparking and meet fire protection requirements.

[0004] However, when these existing technologies are applied to indoor construction vehicles, their limitations gradually become apparent. Indoor construction vehicles, such as trucks, forklifts, cranes, and excavators, often need to operate in a variety of complex and delicate environments, such as food processing plants, lumber mills, and convention centers. These scenarios place higher demands on vehicle stability and maneuverability, as the vehicles must not only avoid damaging the ground but also ensure safe driving on various surfaces to prevent collisions caused by slipping.

[0005] In practical applications, indoor engineering vehicles often encounter track deviations caused by the rubber tracks on both sides running on surfaces with different friction coefficients. For example, one side may be a rough concrete surface with a higher friction coefficient, while the other side may be a smooth steel plate with a lower friction coefficient. Furthermore, indoor surfaces may be subject to localized oil or water stains, causing the rubber tracks to run on surfaces with different friction coefficients. In these situations, due to the characteristics of the surface material and the rubber track itself, rubber tracks struggle to maintain the same grip on different surface materials through their own weight and structure as steel tracks do. This can easily lead to track deviations, such as slipping. This slipping is particularly dangerous in indoor environments, where space is relatively confined and the vehicle is in close proximity to surrounding objects or people. Once slipping occurs, it is very likely to cause a collision with surrounding objects or people, causing serious damage and safety hazards.

[0006] While existing technologies offer some level of contact patch and tension adjustment, these adjustments often rely on pre-set patterns or manual intervention, making it impossible to dynamically adjust the contact patch and tension of the rubber track in real time based on the surface material. This makes it difficult to meet the high stability and maneuverability requirements of engineering vehicles under complex and ever-changing indoor working conditions. In particular, for some engineering vehicles operated remotely or automatically, a track system lacking real-time dynamic adjustment capabilities can cause vehicle instability during driving, impacting operational efficiency and safety. Summary of the Invention

[0007] In view of the deficiencies in the prior art, the present invention aims to provide a method for controlling a rubber tracked indoor engineering vehicle to solve the problems mentioned in the background technology section above.

[0008] The present invention is achieved through the following technical solutions:

[0009] A method for controlling a rubber tracked indoor engineering vehicle, the method comprising the following steps:

[0010] S1. Obtain a preset walking trajectory through path planning of the engineering vehicle, and obtain the actual walking trajectory through an inertial measurement unit;

[0011] S2, comparing the preset walking trajectory with the actual walking trajectory to determine whether the trajectory deviation exceeds a preset threshold, and if so, proceeding to step S3;

[0012] S3. Scan the ground ahead using a multimodal sensor to obtain a ground feature vector, match the ground feature vector with a pre-established material and friction coefficient matching model, and determine the friction coefficient of each of the rubber tracks on both sides.

[0013] S4. Adjust the tension or contact area of the rubber tracks on both sides according to the friction coefficient to reduce the track deviation to below the preset threshold.

[0014] Furthermore, in step S3, the ground ahead is scanned by a multimodal sensor to obtain a ground feature vector, and the ground feature vector is matched with a pre-established material and friction coefficient matching model, specifically including:

[0015] Scanning the ground ahead using a multimodal sensor to acquire first reflection characteristic data and first texture distribution data, the multimodal sensor including a laser sensor and an image sensor, the first reflection characteristic data being generated by the laser sensor, and the first texture distribution data being generated by the image sensor, thereby obtaining a raw data set including signal intensity and surface roughness;

[0016] According to the first reflection feature data and the first texture distribution data, a normalization method is used to process the first reflection feature data, and a denoising method is used to process the first texture distribution data to obtain a second feature data set;

[0017] If the feature dimension of the second feature data set is greater than a preset threshold, a principal component analysis method is used to perform dimensionality reduction processing on the second feature data set, and the reflection feature and texture distribution are integrated to generate a ground feature vector;

[0018] The friction coefficient is obtained by matching the ground feature vector with the pre-established material and friction coefficient matching model.

[0019] Furthermore, in step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0020] If the difference in friction coefficients between the rubber tracks on both sides is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be straight-moving, the tension and contact area of the rubber tracks on both sides are adjusted to a balanced state;

[0021] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0022] Furthermore, in step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0023] If the difference in friction coefficients between the two rubber tracks is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be turning, the tension of the rubber track on one side is adjusted until the track deviation is less than the preset threshold;

[0024] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0025] Furthermore, in step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0026] If the difference in friction coefficients between the rubber tracks on both sides is greater than a preset threshold, the contact area of the rubber track on the side with the lower friction coefficient is increased until the track deviation is less than the preset threshold;

[0027] If the track deviation is still greater than the preset threshold after adjustment, further tightening the tension of the rubber track on the side with the lower friction coefficient until the track deviation is less than the preset threshold;

[0028] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0029] Furthermore, when an abnormal alarm is triggered, the following steps are taken to process the alarm, specifically including:

[0030] Acquiring real-time contact force data of the rubber tracks on both sides, and triggering load eccentricity adjustment if the difference in contact force between the rubber tracks on both sides exceeds a preset threshold, thereby expanding the contact area of the rubber track on the side with greater contact force to balance the contact force distribution of the rubber tracks on both sides and reduce the track deviation to below the preset threshold;

[0031] If the trajectory deviation is still greater than the preset threshold after adjustment, manual intervention is triggered for further processing.

[0032] Furthermore, in step S3, scanning the ground ahead by a multimodal sensor further includes:

[0033] If the ground feature vector obtained by scanning with the multimodal sensor cannot be matched with the pre-established material and friction coefficient matching model, real-time contact force data, driving force data and actual acceleration of the engineering vehicle are obtained to calculate the real-time friction coefficient;

[0034] The real-time friction coefficient is used to adjust the data weight of the multimodal sensor using a gradient descent algorithm to optimize the accuracy of subsequent acquisition of ground material information.

[0035] Furthermore, in step S3, scanning the ground ahead by a multimodal sensor further includes:

[0036] Use multimodal sensors to scan the ground ahead to determine whether there are steps or obstacles on the ground;

[0037] If there is a step, the tension and contact area of the rubber tracks on both sides are adjusted by identifying the height of the step;

[0038] If there is an obstacle, an obstacle avoidance operation is performed to adjust the preset walking trajectory to ensure the safe passage of the engineering vehicle.

[0039] The beneficial effects of the present invention are as follows: the present invention provides a method for controlling rubber tracks in indoor engineering vehicles. By acquiring data from multimodal sensors and track deviation data, the method accurately identifies the ground material and adaptively adjusts the contact area and tension of the rubber tracks to effectively compensate for track deviation. The present invention is capable of sensing changes in ground material in real time and dynamically adjusting the contact area and tension of the rubber tracks accordingly. This technology can not only enhance the adaptability of the vehicle on different ground materials, but also ensure that indoor engineering vehicles travel along a preset walking track, thereby reducing slippage and effectively reducing the risk of collision, thereby improving the overall operating efficiency and safety of indoor engineering vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flow chart of a method for controlling a rubber track on an indoor engineering vehicle. DETAILED DESCRIPTION

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. It should be noted that the description of these embodiments is intended to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0042] Reference Figure 1 As shown, a method for controlling a rubber tracked indoor engineering vehicle comprises the following steps:

[0043] S1. Obtain a preset walking trajectory through path planning of the engineering vehicle, and obtain the actual walking trajectory through an inertial measurement unit;

[0044] S2, comparing the preset walking trajectory with the actual walking trajectory to determine whether the trajectory deviation exceeds a preset threshold, and if so, proceeding to step S3;

[0045] S3. Scan the ground ahead using a multimodal sensor to obtain a ground feature vector, match the ground feature vector with a pre-established material and friction coefficient matching model, and determine the friction coefficient of each of the rubber tracks on both sides.

[0046] S4. Adjust the tension or contact area of the rubber tracks on both sides according to the friction coefficient to reduce the track deviation to below the preset threshold.

[0047] Specifically, the path planning module analyzes the indoor engineering vehicle's route and generates preset trajectory data. The inertial measurement unit collects the actual trajectory data of the engineering vehicle in real time, generating dynamic path information including position and direction. A comparison is performed between the preset and actual trajectory data to determine the deviation between the two trajectories. If the deviation exceeds a preset threshold, the deviation adjustment process is triggered, and subsequent data processing begins. If it does not, real-time collection of actual trajectory data continues.

[0048] More specifically, the path planning module analyzes the walking routes of indoor engineering vehicles, and the A* algorithm is used in combination with environmental map data to generate preset walking trajectory data, with trajectory accuracy controlled at the centimeter level. The actual walking trajectory data of the engineering vehicle is collected in real time through the inertial measurement unit to obtain dynamic path information including X-axis, Y-axis position and yaw angle, with a sampling frequency of 100Hz. Based on the preset walking trajectory data and the actual walking trajectory data, a comparative calculation is performed, and the Euclidean distance formula is used to calculate the deviation value between the two trajectories, with a threshold value set to 5cm. If the deviation value exceeds 5cm, the deviation adjustment process is triggered and the subsequent data processing link is entered; if it does not exceed, the actual trajectory data continues to be collected in real time.

[0049] The beneficial effects of the above embodiment are: by comparing the preset and actual trajectories to determine deviations, combining multimodal sensors to collect ground feature data, generating a ground feature vector and matching it with the friction coefficient, and then adjusting track parameters in a targeted manner. The tension level or contact area can be adaptively adjusted based on the difference in friction between the two tracks. For example, in the prior art, contact area adjustment can be achieved by controlling the filling and discharge of oil in the double-acting hydraulic cylinder in the contact area adjustment mechanism; and by controlling the tensioning cylinder in the track tensioning mechanism, the rubber track can be tightened and loosened.

[0050] In step S3, the ground ahead is scanned by a multimodal sensor to obtain a ground feature vector, which is then matched with a pre-established material and friction coefficient matching model, specifically including:

[0051] Scanning the ground ahead using a multimodal sensor to acquire first reflection characteristic data and first texture distribution data, the multimodal sensor including a laser sensor and an image sensor, the first reflection characteristic data being generated by the laser sensor, and the first texture distribution data being generated by the image sensor, thereby obtaining a raw data set including signal intensity and surface roughness;

[0052] According to the first reflection feature data and the first texture distribution data, a normalization method is used to process the first reflection feature data, and a denoising method is used to process the first texture distribution data to obtain a second feature data set;

[0053] If the feature dimension of the second feature data set is greater than a preset threshold, a principal component analysis method is used to perform dimensionality reduction processing on the second feature data set, and the reflection feature and texture distribution are integrated to generate a ground feature vector;

[0054] The friction coefficient is obtained by matching the ground feature vector with the pre-established material and friction coefficient matching model.

[0055] The beneficial effects of the above embodiment are: the reflection characteristics of the ground are identified by the laser sensor and the texture distribution characteristics of the ground by the image sensor. Through the collaborative work of the two sensors, the physical and visual characteristics of the ground are captured at the same time, avoiding the limitations of a single sensor, normalizing the reflection characteristics to eliminate dimensional differences, denoising the texture data to improve data quality, and reducing the interference of noise on the classification results. For example, indoor lighting may cause the texture data collected by the image sensor to contain spot noise. The denoising algorithm can retain effective texture features, such as marble cracks, to avoid interference with the classification model; principal component analysis is used to reduce the dimensionality of the second feature data set Processing reduces data redundancy while retaining key information, and at the same time fuses reflection and texture features to generate a compact ground feature vector, reducing the computational complexity of subsequent classification algorithms; based on the reflection characteristics and texture distribution characteristics of pre-constructed different ground surfaces, a pre-established material feature database is prepared, which can specifically cover typical scenarios such as concrete floors, carpets, slippery tiles, wooden floors, and rubber floors. Then, the friction coefficient data of different ground materials are measured experimentally to construct a material and friction coefficient matching model; then, the ground feature vector is quickly matched with the above-established material and friction coefficient matching model to achieve standardization and high scalability of ground material classification.

[0056] As another embodiment, a multimodal sensor can be used to scan the ground ahead to obtain first reflection feature data and first texture distribution data. The first reflection feature data is generated by a laser sensor, and the first texture distribution data is generated by an image sensor, resulting in a raw data set containing signal intensity and surface roughness. Based on the raw data set, the first reflection feature data is normalized to eliminate dimensional differences, resulting in standardized reflection feature data. A denoising algorithm is applied to the first texture distribution data to remove interference factors such as speckle noise, resulting in clear texture feature data. Based on the standardized reflection feature data and clear texture feature data, a second feature data set is constructed to determine a comprehensive description of the ground's physical and visual characteristics. The second feature data set is compared with a pre-established material feature database to determine the ground material classification result. If the comparison indicates that the material classification confidence level is below a set threshold, the multimodal sensor is used to re-collect supplementary data from the ground ahead to obtain new reflection feature and texture feature distribution data. The supplementary data is fused with the original second feature data set to update the comprehensive feature description and determine a more accurate ground material classification result. Then, the friction coefficient data of different ground materials are measured experimentally to construct a material and friction coefficient matching model; then the ground feature vector is quickly matched with the material and friction coefficient matching model established above to achieve standardization and high scalability of ground material classification.

[0057] Specifically, a multimodal sensor scans the ground ahead. The laser sensor generates first reflection feature data, and the image sensor generates first texture distribution data. This results in a raw dataset consisting of signal intensity (e.g., laser reflection intensity of 0.8) and surface roughness (e.g., texture standard deviation of 0.12). Based on this raw dataset, the first reflection feature data is normalized, mapping the reflection intensity values to a range between 0 and 1 to eliminate dimensionality differences. This results in standardized reflection feature data (e.g., normalized reflection intensity of 0.75). A denoising algorithm is applied to the first texture distribution data, using a Gaussian filter to remove speckle noise while retaining valid texture features (e.g., texture standard deviation of 0.10 after denoising), resulting in clear texture feature data. Based on this normalized reflection feature data and the clear texture feature data, a second feature dataset is constructed to determine a comprehensive description of the ground's physical and visual characteristics (e.g., a reflection feature weight of 0.6 and a texture feature weight of 0.4). The second feature dataset is compared with a pre-established database of material features, and the K-nearest neighbor algorithm (K=5) is used to determine the ground material classification result (e.g., classification confidence is 85%). If the comparison result indicates that the material classification confidence falls below the set threshold (e.g., confidence is below 90%), the multimodal sensor re-collects supplemental data from the ground ahead to obtain new reflectance features (e.g., reflectance intensity is 0.78) and texture distribution data (e.g., texture standard deviation is 0.11). This supplemental data is fused with the original second feature dataset, and the weighted average method is used to update the comprehensive feature description (e.g., reflectance feature weight is adjusted to 0.55, texture feature weight is adjusted to 0.45), resulting in a more accurate ground material classification result (e.g., classification confidence is increased to 92%). The ground material classification result is combined with a pre-established ground material and friction coefficient matching model (e.g., marble has a friction coefficient of 0.6) to determine the corresponding material's friction coefficient. Based on the obtained friction coefficient and the real-time monitored track deviation data (such as a deviation of 5 cm), the PID control algorithm is used to calculate the track tension adjustment parameters (such as an adjustment amount of 0.3) and generate drive control instructions.

[0058] In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0059] If the difference in friction coefficients between the rubber tracks on both sides is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be straight-moving, the tension and contact area of the rubber tracks on both sides are adjusted to a balanced state;

[0060] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0061] In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0062] If the difference in friction coefficients between the two rubber tracks is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be turning, the tension of the rubber track on one side is adjusted until the track deviation is less than the preset threshold;

[0063] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0064] In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes:

[0065] If the difference in friction coefficients between the rubber tracks on both sides is greater than a preset threshold, the contact area of the rubber track on the side with the lower friction coefficient is increased until the track deviation is less than the preset threshold;

[0066] If the track deviation is still greater than the preset threshold after adjustment, further tightening the tension of the rubber track on the side with the lower friction coefficient until the track deviation is less than the preset threshold;

[0067] If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

[0068] In summary, when the ground materials on both sides of the rubber tracks are the same, or when the ground materials are different but the friction coefficient is the same or less than the set threshold, such as wood floors (such as red oak floors with a friction coefficient of 0.65) and stone floors (such as marble floors with a friction coefficient of 0.6), the core adjustment parameter is to prioritize the tension of the rubber tracks to solve the driving force transmission problem. When the difference in the friction coefficient of the ground materials on the two sides of the rubber tracks is greater than the set threshold, the priority is to balance the adhesion difference through the contact area.

[0069] The advantage of the above adjustment sequence is that when the friction coefficient of the ground material on both sides of the rubber tracks is the same or less than the set threshold, the friction between the rubber tracks and the drive wheels is directly changed through tension adjustment, quickly correcting the speed difference between the two sides. By balancing the tension, the driving force on both sides is ensured to be symmetrical, avoiding straight-line deviation or uneven power distribution in turns. When the friction coefficient of the ground material on both sides of the rubber tracks is large, the contact area is prioritized to balance the adhesion difference, and the contact area adjustment provides the basic guarantee for adhesion.

[0070] Tension adjustment provides millisecond-to-second feedback on track correction, making it ideal for addressing transient operating conditions. Contact area adjustment provides seconds-to-second feedback on track correction, making it more suitable for addressing persistent adhesion imbalances. The two clearly divide responsibilities, forming a dual-tier control architecture of "rapid correction + long-term protection."

[0071] The above scenario-based priority strategy not only enables the rapidity of tension adjustment to cope with dynamic changes, but also ensures long-term stability on complex surfaces through ground contact area optimization. This combination provides engineering vehicles with an efficient, low-consumption, and highly robust motion control solution.

[0072] When the tension and contact area are adjusted to the maximum rated values and the track deviation problem is still not solved, there is a possibility that the load center of gravity of the engineering vehicle is offset. That is, if the track deviation is still greater than the preset threshold after adjustment, an abnormal alarm is triggered to prompt further processing.

[0073] When an abnormal alarm is triggered, it is processed through the following steps:

[0074] Acquiring real-time contact force data of the rubber tracks on both sides, and triggering load eccentricity adjustment if the difference in contact force between the rubber tracks on both sides exceeds a preset threshold, thereby expanding the contact area of the rubber track on the side with greater contact force to balance the contact force distribution of the rubber tracks on both sides and reduce the track deviation to below the preset threshold;

[0075] If the trajectory deviation is still greater than the preset threshold after adjustment, manual intervention is triggered for further processing.

[0076] Specifically, contact force data from the rubber tracks on both sides is collected using pressure sensors at a sampling frequency of 100Hz. For example, if the pressure sensor collects a pressure value of 1200N on the left track and a pressure value of 800N on the right track, the real-time pressure distribution values of the two tracks are obtained. Based on the acquired contact force data of the two rubber tracks, the pressure difference between the two sides is calculated. For example, the pressure of 1200N on the left minus the pressure of 800N on the right side is 400N. The difference is then determined to see if it exceeds a preset threshold of 200N. If the pressure difference between the two sides exceeds the preset threshold, the load eccentricity adjustment mechanism is triggered, and the position of the rubber track on the side with greater contact force is determined to be the left side. Based on the determined position of the rubber track on the side with greater contact force, the expanded contact area data is calculated to be 0.8 square meters. Based on the adjusted contact area data, real-time contact force data from the two rubber tracks is recollected. For example, the pressure on the left side is reduced to 1000N and the pressure on the right side is increased to 900N. The pressure distribution is determined to be balanced and the track deviation is determined to have dropped below the preset threshold. If the track deviation problem is still not resolved, manual intervention is triggered for further processing.

[0077] In step S3, scanning the ground ahead by a multimodal sensor also includes:

[0078] If the ground feature vector obtained by scanning with the multimodal sensor cannot be matched with the pre-established material and friction coefficient matching model, real-time contact force data, driving force data and actual acceleration of the engineering vehicle are obtained to calculate the real-time friction coefficient;

[0079] The real-time friction coefficient is used to adjust the data weight of the multimodal sensor using a gradient descent algorithm to optimize the accuracy of subsequent acquisition of ground material information.

[0080] The beneficial effect of the above embodiment is that if the ground feature vector cannot be successfully matched with the pre-established material and friction coefficient matching model, the real-time contact force data, driving force data and actual acceleration of the engineering vehicle are obtained to calculate the real-time friction coefficient.

[0081] Specifically, using real-time dynamic data, the system infers the actual friction coefficient from driving force, acceleration, and contact force. This can address issues such as image sensor misjudgment caused by light interference and laser sensor misjudgment caused by mechanical vibration. The gradient descent algorithm dynamically adjusts the weights of the laser and image sensors, prioritizing high-reliability data, and increases ground classification accuracy to over 95%. Examples are as follows:

[0082] First, lighting changes: Warehouse ceiling lights reflect strongly on oily or water-stained areas, interfering with the image sensor's texture recognition. The image sensor mistakenly identifies the floor as "dry epoxy resin" due to the oily reflections, while the laser sensor recognizes the presence of water or oil stains. In this case, the laser sensor's weight is increased.

[0083] Second, mechanical vibration: When a vehicle passes through ground joints, high-frequency jitter is generated, causing instantaneous fluctuations in the laser sensor data. At this time, the accuracy of ground classification can be improved by adjusting the image sensor.

[0084] By obtaining the real-time contact force data, driving force data and actual acceleration of the engineering vehicle, the real-time friction coefficient is calculated using the following formula:

[0085]

[0086] Establishing the loss function ,The sensor data weights are adjusted by the gradient descent algorithm so that the predicted friction coefficient is close to the actual value.

[0087] In step S3, scanning the ground ahead by a multimodal sensor also includes:

[0088] Use multimodal sensors to scan the ground ahead to determine whether there are steps or obstacles on the ground;

[0089] If there is a step, the tension and contact area of the rubber tracks on both sides are adjusted by identifying the height of the step;

[0090] If there is an obstacle, an obstacle avoidance operation is performed to adjust the preset walking trajectory to ensure the safe passage of the engineering vehicle.

[0091] Specifically, a multimodal sensor scans the ground ahead, acquiring raw data containing laser reflection features and image texture distribution. Based on this raw data, the laser reflection features are normalized and the image texture data is denoised, resulting in a processed, clear feature dataset.

[0092] The processed feature data set is compared with the pre-built ground feature database to determine the feature classification results of whether there are steps or obstacles on the ground ahead.

[0093] If a step is detected, a laser sensor measures the height of the step to determine its specific height. Based on this height, combined with a pre-set height and track parameter matching model, the system determines the adjustment parameters for the rubber track tension and contact patch. Based on these adjustment parameters, control commands are generated and transmitted to the track drive system, dynamically adjusting the tension and contact patch.

[0094] If an obstacle is detected, the image sensor captures the obstacle's contour and location data, determining its spatial distribution. Based on this spatial distribution information, combined with the vehicle's real-time location and pre-set path, the system calculates an obstacle avoidance path and determines the new driving direction and speed parameters. The calculated obstacle avoidance path parameters are then used to generate navigation control commands and transmit them to the vehicle control system, completing path adjustments and driving operations.

[0095] Specifically, a multimodal sensor scans the ground ahead, acquiring raw data containing laser reflection features and image texture distribution. The laser sensor captures reflection intensity in the range of 0-1000, and the image sensor resolution is 1920×1080 pixels. Based on the acquired raw data, the laser reflection features are normalized using the Min-Max normalization method to map the reflection intensity values to the range of 0-1. The image texture data is also denoised using a Gaussian filter to remove speckle noise, resulting in a processed, clear feature dataset. This processed feature dataset is then compared with a pre-built ground feature database, and feature similarity is calculated using the K-nearest neighbor algorithm to determine the presence of a step or obstacle on the ground ahead. If a step is detected, the laser sensor further measures the step's height, calculating it using triangulation. The specific step height is determined to be 15 cm. Based on this step height value, combined with a pre-set height and track parameter matching model, linear interpolation is used to determine the rubber track tension to be 80 N and the contact patch to be adjusted to 0.5 square meters. Based on the determined adjustment parameters, control instructions are generated and transmitted to the track drive system. A PID control algorithm is used to dynamically adjust the tension and contact area. If an obstacle is detected, the image sensor acquires the obstacle's contour and position data, and an edge detection algorithm is used to extract the obstacle's outline and determine its spatial distribution. Based on this spatial distribution information, combined with the engineering vehicle's real-time position and preset path, the A* algorithm is used to calculate the obstacle avoidance path plan. The new driving direction is determined to be a 30-degree left turn, with a speed parameter of 0.5 meters per second. The calculated obstacle avoidance path parameters are used to generate navigation control instructions and transmit them to the engineering vehicle control system. A fuzzy control algorithm is used to complete path adjustment and driving operations.

[0096] Multimodal sensors are used to scan the ground ahead to determine whether there are steps or obstacles on the ground. This is especially true for some engineering vehicles that are remotely controlled or work automatically. The track system with real-time dynamic adjustment capabilities can prevent the engineering vehicles from becoming unstable during driving, which may affect work efficiency and safety.

[0097] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0098] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, features specified as "first" or "second" may explicitly or implicitly include one or more of the features.

[0099] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for controlling a rubber tracked indoor engineering vehicle, characterized in that: The control method comprises the following steps: S1. Obtain a preset walking trajectory through path planning of the engineering vehicle, and obtain the actual walking trajectory through an inertial measurement unit; S2, comparing the preset walking trajectory with the actual walking trajectory to determine whether the trajectory deviation exceeds a preset threshold, and if so, proceeding to step S3; S3. Scan the ground ahead using a multimodal sensor to obtain a ground feature vector, match the ground feature vector with a pre-established material and friction coefficient matching model, and determine the friction coefficient of each of the rubber tracks on both sides. S4. Adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient to reduce the track deviation to below the preset threshold; If the trajectory deviation after adjustment is still greater than the preset threshold, an abnormality alarm is triggered to prompt further processing. When the abnormality alarm is triggered, the following steps are taken to process the abnormality, including: Acquiring real-time contact force data of the rubber tracks on both sides, and triggering load eccentricity adjustment if the difference in contact force between the rubber tracks on both sides exceeds a preset threshold, thereby expanding the contact area of the rubber track on the side with greater contact force to balance the contact force distribution of the rubber tracks on both sides and reduce the track deviation to below the preset threshold; If the trajectory deviation is still greater than the preset threshold after adjustment, manual intervention is triggered for further processing.

2. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S3, the ground ahead is scanned by a multimodal sensor to obtain a ground feature vector, which is then matched with a pre-established material and friction coefficient matching model, specifically including: Scanning the ground ahead using a multimodal sensor to acquire first reflection characteristic data and first texture distribution data, the multimodal sensor including a laser sensor and an image sensor, the first reflection characteristic data being generated by the laser sensor, and the first texture distribution data being generated by the image sensor, thereby obtaining a raw data set including signal intensity and surface roughness; According to the first reflection feature data and the first texture distribution data, a normalization method is used to process the first reflection feature data, and a denoising method is used to process the first texture distribution data to obtain a second feature data set; If the feature dimension of the second feature data set is greater than a preset threshold, a principal component analysis method is used to perform dimensionality reduction processing on the second feature data set, and the reflection feature and texture distribution are integrated to generate a ground feature vector; The friction coefficient is obtained by matching the ground feature vector with the pre-established material and friction coefficient matching model.

3. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes: If the difference in friction coefficients between the rubber tracks on both sides is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be straight-moving, the tension and contact area of the rubber tracks on both sides are adjusted to a balanced state; If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

4. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes: If the difference in friction coefficients between the two rubber tracks is less than a preset threshold, and the behavior mode of the engineering vehicle is determined to be turning, the tension of the rubber track on one side is adjusted until the track deviation is less than the preset threshold; If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

5. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S4, adjusting the tension or contact area of the rubber tracks on both sides according to the friction coefficient specifically includes: If the difference in friction coefficients between the rubber tracks on both sides is greater than a preset threshold, the contact area of the rubber track on the side with the lower friction coefficient is increased until the track deviation is less than the preset threshold; If the track deviation is still greater than the preset threshold after adjustment, further tightening the tension of the rubber track on the side with the lower friction coefficient until the track deviation is less than the preset threshold; If the trajectory deviation is still greater than the preset threshold after adjustment, an abnormality alarm is triggered to prompt further processing.

6. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S3, scanning the ground ahead by a multimodal sensor also includes: If the ground feature vector obtained by scanning with the multimodal sensor cannot be matched with the pre-established material and friction coefficient matching model, real-time contact force data, driving force data and actual acceleration of the engineering vehicle are obtained to calculate the real-time friction coefficient; The real-time friction coefficient is used to adjust the data weight of the multimodal sensor using a gradient descent algorithm to optimize the accuracy of subsequent acquisition of ground material information.

7. The method for controlling a rubber track indoor engineering vehicle according to claim 1, characterized in that: In step S3, scanning the ground ahead by a multimodal sensor also includes: Use multimodal sensors to scan the ground ahead to determine whether there are steps or obstacles on the ground; If there is a step, the tension and contact area of the rubber tracks on both sides are adjusted by identifying the height of the step; If there is an obstacle, an obstacle avoidance operation is performed to adjust the preset walking trajectory to ensure the safe passage of the engineering vehicle.

Images