A new energy battery potting anti-collision control method and system

Abstract

The application discloses a new energy battery pouring anti-collision control method and system, and relates to the technical field of new energy battery manufacturing automation. The method uses an eye-in-hand vision system installed at the end of the Z-axis to fit a dynamic reference plane to eliminate assembly tolerance. In the path search stage, the application constructs an asymmetric weight matrix based on adaptive adjustment of the dynamic viscosity of glue, and forcibly locks the Z-axis height. At the same time, fluid dynamics pre-check is introduced, the instantaneous angular velocity of the bottom interpolator is extracted, the preset glue rheology database is combined, the lateral spreading amount of the centrifugal force on the glue width is calculated in real time, and the planning speed is automatically limited. When local deadlock triggers obstacle avoidance lifting, the millisecond glue breaking-displacement-glue connecting timing containing negative pressure back absorption and advance pressure building is strictly executed. Through multi-dimensional fluid-motion coupling solution, the application effectively solves the glue throwing distortion caused by high-speed obstacle avoidance in the complex battery pack and the glue breaking and wire drawing problem when obstacles are crossed, and ensures the high consistency of the pouring process.

Description

Technical Field

[0001] This invention relates to the field of automation technology in new energy battery manufacturing, and in particular to a method for anti-collision control of the potting trajectory for complex internal structures of power battery packs. Background Technology

[0002] In the production of new energy batteries, the potting of the battery pack is a critical process that directly affects the battery's heat dissipation performance and insulation safety. The internal structure of a battery pack is extremely complex, densely packed with numerous cell terminals, busbars, separators, and reinforcing ribs, creating a dense obstacle environment. Given the compact internal structure of new energy battery packs and the abundance of terminals and busbars, traditional robotic arms frequently raise the Z-axis to avoid these obstacles, easily leading to adhesive strip stringing or air bubbles being introduced, severely impacting the battery's insulation performance.

[0003] Current battery encapsulation methods primarily rely on XYZ three-axis robotic arms, employing manual teaching to plan trajectories. However, this traditional approach suffers from significant technical limitations:

[0004] 1. Inability to accommodate assembly tolerances: The battery pack tray has tolerances during manufacturing and assembly, and its placement during loading may be tilted. The fixed-track teaching method cannot detect this "dynamic error," which can easily cause the metal potting valve to accidentally collide with the battery terminals during movement, resulting in serious safety accidents such as short circuits, fires, or even explosions.

[0005] 2. Z-axis lifting leads to decreased potting quality: Traditional three-axis equipment must significantly lift the Z-axis to overcome obstacles. This frequent up-and-down movement of the Z-axis disrupts fluid continuity, making the adhesive strip prone to stringing, breakage, or the introduction of air bubbles, severely affecting the consistency and insulation performance of the potting.

[0006] 3. High-speed obstacle avoidance leads to adhesive line distortion: Even with the introduction of R-axis rotation obstacle avoidance, if hydrodynamic factors are not considered, the centrifugal force generated by the rapid rotation of the robotic arm end will throw the adhesive off the trajectory, resulting in uneven adhesive width.

[0007] Therefore, there is an urgent need for an intelligent planning method that can automatically sense assembly tolerances, flexibly avoid obstacles without changing the glue pouring height as much as possible, and overcome hydrodynamic interference. Summary of the Invention

[0008] The purpose of this invention is to provide a new energy battery potting anti-collision control method and system, which aims to solve the problems of glue breakage caused by frequent Z-axis lifting, collision risks caused by assembly tolerances, and glue line distortion during high-speed obstacle avoidance in the prior art.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A collision avoidance control method for the encapsulation trajectory of new energy batteries, the core of which lies in the deep coupling of "process quality" and "motion planning". This invention is no longer a simple geometric path planning method, but a reverse motion solution method based on fluid process constraints. Specifically, it includes:

[0011] 1. Dynamic Baseline Construction: To address the random tilting and manufacturing tolerances during battery pack tray loading, an eye-in-hand vision system scans the bottom surface of the tray and fits a dynamic baseline plane. Instead of relying on an absolute world coordinate system, the system establishes a relative coordinate system based on the workpiece itself, ensuring that the planned path always conforms to the actual posture of the battery pack.

[0012] 2. Rigid Height Locking and Asymmetric Weighting: The traditional "shortest path first" principle is abandoned in favor of a "process first" principle. When searching for a path in the configuration space (C-Space), an asymmetric weight matrix is ​​constructed, assigning a very high penalty weight to changes in Z-axis height (e.g., more than 20 times the weight of the R-axis). This forces the algorithm to search for an R-axis rotation solution (sideways avoidance) when encountering obstacles, only considering Z-axis elevation when no solution is found, thus maximizing the continuity of the adhesive strip.

[0013] 3. Pre-processing verification using fluid dynamics: Unlike existing methods that "plan the path first, then adjust the speed," this invention uses the rheological properties of the adhesive as input parameters for path search. At each node expansion of the search tree, the adhesive rheological model is dynamically input to calculate the effect of centrifugal force on the adhesive width. If the predicted adhesive width distortion exceeds the tolerance, the path node is invalidated or forced to slow down.

[0014] 4. Hierarchical obstacle avoidance strategy: Introducing a local obstacle density discrimination mechanism, adopting a "posture adjustment (sideways movement)" strategy for sparse obstacle areas and a "height adjustment (crossing)" strategy for dense obstacle areas to achieve human-like intelligent decision-making.

[0015] Compared with the prior art, the present invention has the following significant advantages:

[0016] A qualitative leap in process quality: By strictly limiting the Z-axis lift, defects such as stringing, glue breakage, and air bubbles, which are common in traditional obstacle avoidance methods, have been eliminated.

[0017] Extremely high motion stability: The fluid dynamics compensation mechanism solves the problem of uniform adhesive application during high-speed rotation obstacle avoidance, and avoids product scrap caused by centrifugal adhesive spillage.

[0018] Zero tolerance sensitivity: Dynamic reference technology significantly reduces the equipment's requirements for feeding accuracy, thus reducing reliance on precision tooling. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall three-dimensional structure of the dispensing machine according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the internal motion module structure according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the exploded structure of the head assembly according to an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of the battery pack tray structure and potting trajectory according to an embodiment of the present invention;

[0023] Figure 5 This is a logic block diagram of the anti-collision control principle in an embodiment of the present invention (Note: Each functional module in this diagram corresponds to a specific execution unit in the control system).

[0024] Figure 6 This is a flowchart of the anti-collision control process according to an embodiment of the present invention;

[0025] Figure 7 This is a cross-sectional schematic diagram of the obstacle avoidance principle according to an embodiment of the present invention;

[0026] Figure 8 This is a schematic diagram illustrating the generation of a virtual bounding box according to an embodiment of the present invention;

[0027] Figure 9 This is a weighted adaptive relationship curve diagram of an embodiment of the present invention;

[0028] Figure 10 This is a block diagram of the call logic of the rheological database and control system according to an embodiment of the present invention;

[0029] Figure 11 This is a timing waveform diagram of the "glue breaking-shifting-glue joining" action in an embodiment of the present invention.

[0030] Explanation of key component symbols in the diagram: 1-Frame, 2-Display, 3-Loading / unloading platform, 4-Keyboard and mouse placement box; 10-Pouring valve, 11-3D camera; 20-X / Z axis motion assembly, 21-Y axis motion assembly, 22-3D detection assembly, 23-Pouring box, 24-Equipment base plate; 30-Servo motor, 31-Reducer, 32-Sensing plate, 33-Sensor, 34-Rotating assembly, 35-Protective cover; 40-Battery pack, 41-Normal potting trajectory, 42-Avoidance trajectory, 43-Obstacle (virtual bounding box). Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0032] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0033] I. Hardware System Implementation Examples

[0034] like Figures 1 to 3 As shown, the intelligent potting device provided by the present invention is based on a hardware architecture that is "eye in hand" and "multi-axis flexible", providing a physical carrier for subsequent intelligent algorithms.

[0035] 1. Motion execution mechanism:

[0036] like Figure 2 As shown, the equipment adopts a four-axis linkage platform (X / Y / Z / R). The head assembly (...) Figure 3 The integrated rotating component (part 34) allows the 3D camera (part 11) and the potting valve (part 10) mounted on it to not only perform linear interpolation motion along the X / Y / Z axes, but also to rotate ±360° along the R axis. This design enables the 3D camera to adjust its orientation and inspect the internal structure of the product from all angles, avoiding the blind spots of a fixed camera.

[0037] 2. Visual perception system:

[0038] like Figure 3 As shown, an eye-in-hand architecture is adopted, with the 3D camera (component 11) rigidly fixed to the Z-axis motion module via a connecting bracket. Compared to a fixed camera, this mounting method can perform multi-angle, high-precision scanning of local details of the battery pack and can compensate for parallax caused by Z-axis motion in real time. The 3D camera is preferably a high-precision line laser profilometer or a structured light camera.

[0039] 3. Target audience:

[0040] like Figure 4 As shown, the new energy battery pack tray (component 40) has several partitions / reinforcing ribs and poles inside, which are marked as "static obstacles" in the subsequent algorithm.

[0041] II. Examples of Collision Avoidance Control Methods

[0042] like Figure 5 , Figure 6 and Figure 8 As shown, the collision avoidance control method of the present invention is executed by the collision avoidance control engine in the control system. In order to ensure the accuracy of path planning and process adaptability, the core algorithm steps are broken down as follows in this embodiment:

[0043] Step S1: Baseline Construction and Dynamic Compensation. In actual production, battery pack conveyor belts often experience vibrations, resulting in translational deviations of Δx and Δy and rotational deviations of Δθ when the pallet arrives at the workstation. This step aims to establish a relative coordinate system that varies with the workpiece.

[0044] S1-1 Data Acquisition: Control the Z-axis motion module to drive the 3D camera (part 11) to perform multi-point or continuous scanning above the battery pack tray to acquire the raw point cloud dataset. .

[0045] S1-2 Plane Fitting: The RANSAC (Random Sample Consensus) algorithm was used for... The system performs processing. It randomly selects three points from the point cloud to construct a planar model, calculates the distances from the remaining points to this plane, and iterates through the process. Second-rate( (≥100) to remove noise, and finally fit the mathematical equation x+By+Cz+D=0 for the bottom plane of the tray.

[0046] S1-3 Coordinate System Transformation: Construct a "dynamic workpiece coordinate system" based on the fitted plane equations. This involves transforming the absolute coordinates of all subsequent obstacles. Converted to relative coordinates relative to this reference plane Technical effect: Regardless of the physical tilt angle of the tray, the system always uses z'=0 as the bottom reference, ensuring that the glue filling height H (e.g., H=3mm) remains constant relative to the bottom surface of the battery pack, thereby completely eliminating physical assembly tolerances.

[0047] Step S2: Obstacle Layer Extraction and Modeling

[0048] S2-1 Height Threshold Segmentation: The algorithm sets a height threshold based on the aforementioned dynamic reference plane. (For example, 3mm above the reference plane). Traverse the point cloud data, and... The point clusters are marked as potential obstacles.

[0049] S2-2 Bounding Box Generation: Euclidean clustering is performed on the marked point clusters to identify independent objects such as poles and partitions. For example... Figure 8 As shown, a minimum bounding rectangle or cylinder, i.e., a red bounding box (part 43), is generated for each object in the virtual environment (digital twin model).

[0050] S2-3 Expansion Treatment: Taking into account the physical radius of the dispensing valve and safety margin The bounding box is scaldned to generate configuration space obstacles (C-Obstacles) for pathfinding.

[0051] Step S3: Multidimensional Coupled Path Search and Fluid Control (Core Innovative Step) This step searches for a collision-free path between the starting and ending points. To address the "glue breakage" and "glue spillage" problems, this invention introduces asymmetric weighting and a pre-verification mechanism based on fluid dynamics.

[0052] S3-1 Constructing an Asymmetric Weighted Cost Function and Adaptive Mechanism

[0053] In complex real-world scenarios, the weights of the cost function cannot be fixed values. This invention introduces a dynamic weight adaptive allocation mechanism. Specifically, the total path cost J is defined as the weighted sum of all dimensions.

[0054] The system establishes the following parameter mapping relationship:

[0055] (1) Basic displacement weight The constant value is 1.0, representing the basic cost of spatial distance.

[0056] (2) Attitude adjustment weights The value ranges from 1.2 to 5.0. This applies when the visual perception shows a small gap between obstacles. Approaching the lower limit encourages the algorithm to perform lateral rotation to avoid obstacles.

[0057] (3) Z-axis height locking weight The system is highly sensitive to the viscosity of the adhesive solution, and has a pre-set adaptive formula. ,in This refers to the dynamic viscosity of the adhesive currently in use. This is the standard reference viscosity. It is not a fixed constant, but is determined by the calibration program of the underlying control system through trial extrusion and measurement by calibration sensors during equipment initialization or when changing the glue type.

[0058] The physical meaning of this setting is that the higher the viscosity of the adhesive (the easier it is to form strings), the more linearly the penalty weight for Z-axis lifting is amplified, especially under extreme conditions. Reachable More than 50 times, completely blocking unnecessary Z-axis elevation from the underlying algorithm.

[0059] S3-2 Fluid Dynamics Pre-verification and Rheological Database

[0060] This invention deeply integrates the rheological properties and kinematics of adhesives. A two-dimensional adhesive rheological database (LUT lookup table) is pre-installed in the system's non-volatile memory. As shown in Table 1, the database stores key physical parameters using "adhesive type" and "operating temperature" as a joint primary key.

[0061] Table 1: Glue Rheology Database (Example Excerpt) Adhesive type Operating temperature (°C) Density p (g / cm³) before curing Dynamic viscosity u (mPa·s) Experience broadening factor k Thermally conductive silicone-A 20 2.15 8500 0.042 Thermally conductive silicone-A 30 2.12 7200 0.048 Polyurethane-B 25 1.65 4500 0.085

[0062] The verification execution logic is as follows:

[0063] When expanding a specific node, the underlying trajectory planner uses calculus to derive the instantaneous angular velocity by looking ahead to adjacent interpolation points. The system automatically retrieves Table 1. , , .

[0064] System extracts angular velocity Then, the empirical calculation model for lateral widening at the bottom layer of the control system is directly invoked:

[0065]

[0066] In the formula: To predict the distortion variable for glue width, The density of the adhesive solution, Let ω be the instantaneous angular velocity, r be the radius of gyration, and h be the dynamic glue pouring height. For dynamic viscosity, The comprehensive broadening factor is obtained by fitting through a large number of fluid spinning experiments.

[0067] This physical model reveals that centrifugal force and height are positively correlated with broadening, while viscosity plays a role in resisting deformation damping. The system will predict... With process tolerances (such as) (Compare.) If the error exceeds the tolerance, the algorithm triggers a bisection method to adjust the angular velocity at the adjustment point. The verification process is iterated again until the requirements are met, thus preventing "glue slippage" before physical execution.

[0068] S3-3 Deadlock Relief and Hard Real-Time Timing for "Glue Disconnection-Displacement-Glue Connection"

[0069] When excessive local obstacle density causes full-angle deadlock on the R-axis, the system releases the Z-axis lock. To prevent forced crossing from causing stringing or air bubbles, this invention employs millisecond-level hard real-time control timing (executed in conjunction with a fieldbus):

[0070] time (Trigger): Issue an emergency stop and deceleration command, and each axis smoothly decelerates to 0 within 50ms.

[0071] time (Glue Cut-off): Output valve-closing level and forcibly superimpose a 150ms negative pressure suck-back pulse to pull back the residual glue at the end of the nozzle and cut off the stringing.

[0072] time (Crossing): The Z-axis rapidly rises to the safe crossing height with an acceleration of 0.5G. Interpolate the X and Y axes to the landing joint point.

[0073] time (Descent): The Z-axis descends to return to the working height.

[0074] time (Pre-build pressure): To prevent insufficient rubber at the start (slimming effect), build pressure in advance. (e.g., 120ms) Open the glue valve to perform pre-extrusion and pressure building. The moment the pressure building in the pipeline cavity is completed, the multi-axis linkage interpolation seamlessly recovers.

[0075] S3-4 Path Generation: Based on the above weighting and verification mechanisms, the algorithm automatically generates a trajectory with "unchanged Z-axis height, R-axis rotation for avoidance," and "velocity adapted to centrifugal force" (e.g., ...). Figure 7 (As shown in trajectory 42).

[0076] Step S4: Command Issuance and Physical Driver Execution:

[0077] S4-1 Control Command Conversion: Converts the generated safe path trajectory data into G-code or low-level pulse control signals that can be recognized by the CNC system.

[0078] S4-2 Hardware Co-drive: The above pulse signals are sent to the servo driver of the multi-axis motion platform, and the start and stop timing of the glue valve is converted into the level trigger signal of the I / O port, strongly coupling the physical space displacement of the multi-axis mechanical mechanism with the opening and closing of the glue valve.

[0079] S4-3 Closed-Loop Monitoring Execution: During physical execution, the motor torque is monitored in real time. If an abnormal collision torque is detected, the emergency stop braking hardware is immediately triggered.

[0080] To verify the effectiveness of the "asymmetric weight height locking" and "fluid dynamics pre-verification" described in this invention in actual industrial settings, the applicant relied on a real four-axis dispensing device equipped with the system of this invention (such as...). Figure 1 A systematic comparative experiment was conducted (as shown in the figure).

[0081] 1. Experimental conditions and test subjects:

[0082] A certain type of new energy storage battery pack contains 32 cylindrical battery cells and staggered busbars, with obstacle heights ranging from 5mm to 25mm.

[0083] Adhesive used: Two-component thermally conductive structural polyurethane adhesive (initial dynamic viscosity) ).

[0084] Target dispensing parameters: Design width of dispensing strip Tolerance requirements The target running linear speed is 80 mm / s.

[0085] 2. Experimental Grouping

[0086] Control group (traditional process): Uses the conventional shortest path planning algorithm of a three-axis robotic arm, and performs Z-axis lifting to cross obstacles (Z-axis obstacle avoidance) without a fluid verification mechanism.

[0087] Experimental group (the present invention scheme): Enable asymmetric weight matrix (force Z-axis locking), prioritize R-axis attitude side-walking obstacle avoidance, and enable fluid pre-verification dynamic speed limit.

[0088] 3. Comparison of Experimental Data and Indicators: For each group, 500 battery packs were continuously encapsulated. The following key process indicators were statistically analyzed using a 3D vision online inspection system:

[0089] detection indicators Control group (traditional Z-axis elevation) Experimental Group (R-axis obstacle avoidance + fluid calibration of this invention) Effect Improvement Analysis Breakage rate (times / 100 pieces) 14.2 times 0 times The defect of the glue line breaking due to frequent Z-axis lifting has been completely eliminated. Maximum error in glue width +1.8mm (glue distortion) +0.25mm Centrifugal force prediction speed limit strictly controls the glue width error within the tolerance. bubble mixing rate 4.8% 0.1% Maintaining a constant glue dispensing height prevents the glue strip from folding and trapping air bubbles in the air. Average beat rate per package (CT) 125 seconds 108 seconds This eliminates the acceleration and deceleration time loss during Z-axis start-stop, resulting in an overall efficiency improvement of 13.6%.

[0090] 4. Experimental Conclusions

[0091] The above-mentioned real bench test data shows that by coupling a rheological physical model in the underlying path search, the present invention not only fundamentally solves the problems of glue breakage and air bubbles that are prone to occur when high-viscosity adhesives cross obstacles, but also overcomes the centrifugal glue-spraying distortion caused by high-speed obstacle avoidance on the R-axis, achieving a double leap in potting quality and production cycle time.

[0092] Through the above-mentioned refined implementation steps, the present invention realizes closed-loop control of the entire process from environmental perception, dynamic benchmark construction, priority attitude obstacle avoidance to fluid process compensation, effectively solving the problems of glue breakage and collision in the glue filling process of new energy batteries.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A new energy battery potting anti-collision control method, characterized in that, Includes the following steps: Step S1: Control the vision system at the Z-axis end of the multi-axis motion platform to acquire point cloud data, fit and generate a dynamic reference plane and establish a relative coordinate system; Step S2: Identify obstacles based on the dynamic reference plane and generate a bounding box model with expansion margin; Step S3: Perform a fluid-motion coupling path search based on the preset potting path. During the search process, "potting process quality" is quantified as a rigid constraint condition, and the following logic is executed: Rigid height locking: Construct an asymmetric weight cost function, set Z-axis displacement weight, R-axis rotation weight and path point displacement weight, and adaptively adjust and amplify the Z-axis displacement weight based on the rheological properties of the current adhesive, and force priority to solve the R-axis attitude obstacle avoidance. Fluid pre-verification: During node expansion, the instantaneous angular velocity of the current planned trajectory is extracted, substituted into the glue rheology model and the predicted width expansion formula, and the glue width distortion corresponding to the centrifugal tangential force is calculated; if the distortion exceeds the tolerance, the planning speed of the node is automatically reduced. Controlled timing of crossing: If the R-axis attitude solution gets stuck in deadlock, the Z-axis lock is released to allow crossing, and the glue breaking action including negative pressure back suction and glue joining action including pre-pressure build-up are inserted strictly according to the preset timing before and after the crossing action. Step S4: The generated instruction set, which includes trajectory coordinates, R-axis attitude and fluid matching speed, is converted into pulse signals to control the underlying motor and level trigger signals to control the glue valve, thereby driving the multi-axis motion platform and glue dispensing execution component to perform anti-collision potting operations in the physical space.

2. The method of claim 1, wherein, In step S3, an asymmetric weighted cost function is used to strengthen the rigid height locking, as detailed below: In the cost function J of path planning, the displacement weight of basic path point is set ; Based on the dynamic viscosity of the adhesive extracted from the control system rheological database Adaptive calculation of Z-axis displacement weights ,in , The preset empirical coefficient is forcibly set by the control system. ; Set R-axis rotation weight satisfy This ensures that the algorithm will inevitably converge to a unique solution in non-dense obstacle areas by rotating along the R-axis to avoid obstacles.

3. The method according to claim 1, characterized in that, The "fluid pre-verification" in step S3 specifically includes: (1) Extracting angular velocity: from the look-ahead interpolation period of the underlying trajectory planner In the process, the instantaneous angular velocity is obtained by differentiating the target attitude angles of adjacent interpolation points. ; (2) Obtaining rheological parameters by looking up a table: Based on the current adhesive type and operating temperature, search the rheological database in the control system to obtain the density before curing. Dynamic viscosity and experience broadening coefficient ; (3) Predicting the distorted variable: Calculating the centrifugal tangential force ; Then, substitute the values ​​into the expansion model to calculate the lateral expansion amount. Where r is the radius of rotation and h is the current dispensing height; (4) Speed ​​correction: If If the predicted width exceeds the set tolerance threshold, the planned linear velocity and angular velocity of the current node will be reduced proportionally until the predicted width returns to the safe range.

4. The method according to claim 1, characterized in that, The timing sequence of the "timing-controlled crossing" glue breaking and glue joining actions in step S3 is as follows: time The motion controller issues an emergency stop command, and the motion mechanism smoothly decelerates to zero at the deadlock point; time : Output a valve-closing command to the dispensing execution component, and simultaneously add a duration of . A millisecond negative pressure back-suction pulse cuts off the fluid drawing process; time The Z-axis is raised to the crossing height with a set safety acceleration, and then the XY-axis plane moves to above the landing adhesive point; time The Z-axis descends to the working height, and the system remains stationary and waiting. time :in advance A valve opening command is sent in milliseconds to perform pre-extrusion pressure build-up. Once the pressure in the pipeline reaches the pressure build-up threshold, the linkage interpolation of each axis is restored.

5. The anti-collision control method for the potting trajectory of a new energy battery according to claim 1, characterized in that, In step S1, the eye-in-hand vision system is directly fixed to the end of the Z-axis and rotates with the R-axis; the method further includes: during the potting process, using the eye-in-hand vision system to monitor the glue-hanging status of the glue nozzle in real time; if the glue-hanging accumulation is detected to exceed the threshold, the safety expansion margin of the bounding box model is dynamically increased in the path planning of step S3 to prevent glue-hanging drips from contaminating the battery terminals.

6. The anti-collision control method for the potting trajectory of a new energy battery according to claim 1, characterized in that, In step S2, the generation of the bounding box model includes an expansion process; the size of the bounding box model is determined based on the sum of the physical contour of the static obstacle, the physical radius of the glue head, and a preset safety clearance margin.

7. The anti-collision control method for the potting trajectory of a new energy battery according to claim 1, characterized in that, Step S3 further includes smoothing the generated path; fitting the polyline segments in the path using the B-spline curve algorithm, and inserting transition points in road segments with large R-axis rotation angle changes to ensure the continuity of the angular velocity of the motion actuator.

8. The anti-collision control method for the potting trajectory of a new energy battery according to claim 1, characterized in that, Step S4 specifically includes: loading the kinematic model of the motion actuator, the glue-filling head model, and the bounding box model in the virtual environment of the controller; performing discretized step simulation along the planned path to detect whether the glue-filling head model and the bounding box model interfere at each step point; if a collision risk is detected, extracting the feature information of the collision area, adjusting the step size parameter or obstacle avoidance weight of the path search algorithm, and returning to step S3 to replan.

9. A new energy battery potting anti-collision control system, used to implement the method described in any one of claims 1 to 8, characterized in that, include: A multi-axis motion actuator, including linear motion components for the X, Y, and Z axes and a rotary component for the R-axis at the end; The eye-in-the-hand visual perception module is rigidly connected to the end of the Z-axis linear motion component and is used to acquire relative point cloud data with the bottom plane of the battery pack as a reference. The glue dispensing execution component is installed adjacent to the eye-in-the-hand visual perception module and is driven by the R-axis rotation component to adjust the glue dispensing angle. The control system is connected to the multi-axis motion execution mechanism, the eye-in-the-hand visual perception module and the glue dispensing execution component respectively.

10. The anti-collision intelligent planning system for the potting trajectory of a new energy battery according to claim 9, characterized in that, The control system is equipped with an anti-collision control module, which is configured to execute a search algorithm based on an asymmetric weight matrix and fluid dynamics constraints. The control system is also configured to output control signals in real time to adjust the dispensing pressure and physical movement speed of the dispensing execution component based on the R-axis angular velocity and centrifugal force verification results of the planned path.

Images