Time-energy dual objective optimization path planning method for bagged material grabbing

Abstract

The application discloses a time-energy dual target optimization path planning method for bagged material grabbing, and relates to industrial robot motion control technology and intelligent warehouse logistics loading and unloading technology. In the method, through the technologies of grabbing configuration constraint modeling, improved RRT path search, 5-order B-spline trajectory smoothing, time-energy dual target trajectory optimization, dynamic obstacle avoidance and real-time trajectory correction, high-precision, high-efficiency, low-energy-consumption and high-robustness path planning of the industrial robot in the bagged material grabbing scene is realized, and the scheme is completely adaptive to the bagged material pose change and the dynamic operation environment, and does not need manual intervention.

Description

Technical Field

[0001] This application relates to industrial robot motion control technology and intelligent warehousing and logistics loading and unloading technology, specifically to an industrial robot path planning method and system for bagged material grasping scenarios, applicable to fast and energy-saving path planning scenarios for bagged material grasping in fully automated industrial loading, unloading and depalletizing scenarios. Background Technology

[0002] Currently, industrial robots are widely used in the automated loading, unloading, and depalletizing of bagged materials. Existing path planning schemes for bagged material handling mainly fall into two categories: one is a teach-based fixed waypoint trajectory playback scheme, where technicians pre-teach the robot the grasping path, and the robot directly playbacks the trajectory during operation; the other is a kinematics-based path planning scheme, such as using standard RRT (Robotic Tracking Method). (Fast traversal of random trees), A algorithm or trapezoidal velocity planning can be used to implement point-to-point path search and velocity planning. Some solutions will combine simple obstacle avoidance rules to complete trajectory generation.

[0003] These methods have the following drawbacks: 1. The fixed waypoint trajectory playback scheme based on teaching is fixed once the path is taught. It cannot adapt to the inter-frame changes in the position and posture of bagged materials in the stack. It is prone to problems such as approach angle deviation leading to bag slippage and collision with adjacent bags, resulting in a low grasping success rate. Moreover, when the stack layout changes, manual re-teaching is required, which is time-consuming and labor-intensive, and cannot achieve unmanned operation.

[0004] 2. Based on standard RRT The path planning scheme uses uniform random sampling, which leads to slow convergence. The joint space Euclidean distance metric does not consider the differences in contribution of each joint to the end pose. The collision detection efficiency is low and the path search takes a long time, making it difficult to meet the real-time requirements of industrial sites.

[0005] 3. Traditional trapezoidal speed planning and A algorithm only consider time efficiency or geometric path optimization, without performing system-level optimization between time efficiency and energy consumption, resulting in high energy consumption. At the same time, they do not take into account special process constraints such as approach angle and load torque for flexible body grasping of bagged materials, and lack the ability to correct trajectory in real time under dynamic environments. When facing dynamic scenarios such as conveyor belt conveying and personnel / forklift intrusion, collisions are likely to occur and the trajectory cannot be adaptively adjusted. Summary of the Invention

[0006] To address the existing problems in path planning during bagged material grasping, this application fully considers the special characteristics of the bagged material grasping scenario and proposes a set of methods including grasping configuration constraint modeling and improved RRT. This method employs a path planning approach that incorporates search, B-spline smoothing, time-energy dual-objective optimization, and dynamic obstacle avoidance with real-time trajectory correction. Addressing the shortcomings of existing bagged material grasping path planning schemes, such as low grasping success rate, inability to balance time and energy consumption, slow convergence speed, and lack of support for dynamic obstacle avoidance, this method improves upon existing approaches by implementing grasping configuration constraint modeling and an improved Real-Time Tracking (RRT). By employing methods such as path search, 5th-order B-spline trajectory smoothing, time-energy dual-objective trajectory optimization, dynamic obstacle avoidance, and real-time trajectory correction, high-precision, high-efficiency, low-energy-consumption, and highly robust path planning is achieved for industrial robots in bagged material grasping scenarios. Moreover, the method is fully adaptive to changes in the pose of the bagged material and the dynamic operating environment, requiring no human intervention.

[0007] On the one hand, embodiments of this application provide a time-energy dual-objective optimized path planning method for bagged material grasping, including: S10, construct the gripping configuration constraints, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints; S20, based on data acquired from multiple sensors, utilizes an improved RRT The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection. S30, the collision-free path is smoothed using a 5th-order B-spline curve to obtain a smooth path; S40, based on the smooth path, establish a dual objective function based on total motion time and total energy consumption, and use the weighted Chebyshev method to convert the dual objective function into a single objective problem. Under the constraints of the grasping configuration, use the sequential quadratic programming method to solve the problem and obtain the optimal path. S50, during the execution of the optimal path, an octree update algorithm is used to maintain the dynamic environment map, and when a new obstacle is detected in the safe corridor, local replanning is performed; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

[0008] On the other hand, embodiments of this application also provide a time-energy dual-objective optimized path planning system for bagged material grasping, including: The grab configuration constraint construction module is used to construct grab configuration constraint conditions, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints. Improved RRT The path search module is used to search for data from multiple sensors, utilizing an improved RRT. The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection. A 5th-order B-spline trajectory smoothing module is used to smooth the collision-free path using a 5th-order B-spline curve to obtain a smooth path. The time-energy dual-objective optimization module is used to establish a dual-objective function based on the total motion time and total energy consumption on the basis of the smooth path. The weighted Chebyshev method is used to transform the dual-objective function into a single-objective problem. Under the constraints of the grasping configuration, the sequential quadratic programming method is used to solve the problem and obtain the optimal path. The dynamic obstacle avoidance and trajectory correction module is used to maintain a dynamic environment map using an octree update algorithm during the execution of the optimal path, and to perform local replanning when a new obstacle is detected in the safe corridor; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

[0009] On the other hand, embodiments of this application also provide a computer storage medium storing a plurality of computer instructions for causing a computer to execute the above-described method.

[0010] The time-energy dual-objective optimization path planning method for bagged material grasping in this application, through the synergistic effect of five improvement points, solves the defects of existing bagged material grasping path planning schemes, achieves multi-dimensional technical improvements, and brings the following beneficial effects: 1. Significantly improved grasping success rate. Through grasping configuration constraint modeling and dynamic trajectory correction, the system can adapt to changes in the position / orientation of bagged material in the stack and adjustments to the stack layout. In 100 grasping tests with different stack layouts, the grasping success rate increased from 72.3% in the traditional teaching scheme to 87.8%, an increase of 15.5 percentage points. Moreover, no manual re-teaching is required, enabling fully automated unmanned operation of bagged material loading and unloading, significantly reducing labor costs.

[0011] 2. Pathfinding efficiency is significantly improved. Improved RRT. The algorithm improves the convergence speed of path search by 65.6% through three aspects, reduces the average number of iterations from 3200 to 1100, and reduces the time of a single path search from 280ms to 95ms, fully meeting the real-time requirements of industrial sites; at the same time, the weighted distance metric shortens the path length by about 18%, further improving the overall operation efficiency.

[0012] 3. Dual optimization of time and energy consumption. By optimizing time and energy consumption in two ways, the limitations of single-objective optimization are broken, and Pareto optimality of the two core indicators is achieved. In the grabbing and placing operation of 50kg bagged material, the cycle time in efficiency-first mode is reduced from 8.2 seconds to 7.1 seconds (saving 13.4%), and energy consumption is reduced from 345 joules to 310 joules (reducing 10.1%). In energy-saving-first mode, energy consumption is reduced to 260 joules (reducing 24.6%). It can be flexibly configured according to the efficiency requirements and energy consumption assessment of the industrial site, and has a high degree of adaptability.

[0013] 4. Smoother robotic arm operation and extended equipment lifespan. The smooth 5th-order B-spline trajectory reduces the maximum joint angular acceleration (jerk) to only 48%~62% of that of the traditional trapezoidal velocity planning scheme, and reduces mechanical impact by 38%~52%. It completely eliminates sudden changes in velocity and acceleration at inflection points, which not only improves the stability of bagged material gripping, but also significantly reduces the wear of core components such as robotic arms and reducers, effectively extending the service life of the equipment.

[0014] 5. Significantly improved operational safety and robustness. Collision avoidance constraints and dynamic obstacle avoidance mechanisms provide dual safety protection. A 20mm safety margin prevents static collisions with adjacent bags and stack frames. 50Hz high-frequency dynamic sensing, local replanning, and emergency smooth braking effectively avoid dynamic collision risks such as conveyor belt feeding and personnel / forklift intrusion. Furthermore, the emergency braking adopts a smooth deceleration method to avoid mechanical damage and bagged material slippage caused by hard stops.

[0015] 6. High path accuracy with no discretization loss. Improved RRT The algorithm performs path search in continuous joint space. Compared with the gridded discrete search of Algorithm A, there is no path accuracy loss problem. The positional accuracy of the end Cartesian space meets the requirements for accurate grasping of bagged materials, and the search time is much lower than that of Algorithm A with high-resolution grid.

[0016] 7. Wide adaptability and strong industrial applicability. The overall method does not depend on specific bag types or robotic arm models, and is suitable for automatic loading, unloading, and depalletizing operations of various bagged materials such as woven bags, kraft paper bags, and composite film bags. It can be directly deployed in existing industrial warehousing and logistics loading and unloading production lines without the need for major modifications to the production lines, resulting in low modification costs and high promotional value. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The flowchart illustrates the time-energy dual-objective optimized path planning method for bagged material grasping provided in this application embodiment.

[0019] Figure 2 This is an architecture diagram of a time-energy dual-objective optimized path planning system for bagged material grasping provided in an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Figure 1 A flowchart illustrating the time-energy dual-objective optimization path planning method for bagged material grasping provided in this application embodiment. This application embodiment provides a time-energy dual-objective optimization path planning method for bagged material grasping, including: S10, construct the gripping configuration constraints, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints.

[0022] For example, the key difference between bagged material gripping and rigid workpiece gripping lies in the flexibility of the bag, which makes the gripping success rate highly sensitive to the approach direction and contact method of the actuator. This application formalizes the process constraints of bagged material gripping as the following mathematical constraints: the approach direction constraint is that the angle between the actuator and the approach direction and the normal vector of the principal plane of the bag is less than or equal to the maximum approach deviation angle; the collision avoidance constraint is that the minimum Euclidean distance between the space occupied by the robotic arm linkage and the actuator assembly and the space occupied by any obstacle is greater than or equal to the safety margin; the joint kinematic constraint is that the angles, angular velocities, and angular accelerations of each joint are all within the hardware allowable range; and the joint torque constraint is that the torque of each joint is less than or equal to the maximum torque that can be withstood.

[0023] Specifically, in the approach direction constraint, let the approach direction vector of the execution end be... (i.e., the z-axis direction of the end coordinate system). Given a 3-dimensional vector space, the normal vector of the principal plane of the bag, obtained by jointly solving the high-precision 3D coordinates and pose of the bagged material from multi-source sensor fusion, is: Then the angle between the approach direction and the normal vector should satisfy:

[0024] in, The superscript T indicates transpose, representing the maximum allowable approach deviation angle. This indicates taking the absolute value. This represents the L2 norm. This constraint ensures that the principal direction of the clamping force is approximately perpendicular to the plane of the bag, maximizing the effective component of the gripping force and preventing the bag from slipping due to excessive tangential force.

[0025] In collision avoidance constraints, the assembly of the robotic arm links and the actuator is assumed to be in a joint configuration. The space occupied below is , For joint space, the first in the environment The space occupied by each obstacle (including adjacent bags, stack frames, conveyor belts, etc.) is Then for any configuration on the trajectory It must meet the following requirements:

[0026] in, Let be the Euclidean distance between the two geometric objects. For safety margins, the geometric representation of environmental obstacles is provided by an octree map constructed from multi-source fused 3D point clouds.

[0027] In the kinematic constraints of the joints, each joint of the robotic arm (each joint corresponds to one degree of freedom, totaling...) The angle, angular velocity, and angular acceleration (each degree of freedom) must be within the limits allowed by the hardware.

[0028] in, , and Let these represent the angle, angular velocity, and angular acceleration of the i-th joint, respectively. and Let be the lower and upper limits of the angle of the i-th joint, respectively. and These are the upper limits of the angular velocity and angular acceleration of the i-th joint, respectively.

[0029] In the joint moment constraints, the load mass of the bagged material is considered. (Typically 20 to 50 kg), the torque of each joint must meet the following requirements:

[0030] in, Let the torque be the torque of the i-th joint. This represents the upper limit of the torque at the i-th joint. The joint torque is calculated using the inverse dynamics model of the robotic arm (Newton-Euler recursive formula):

[0031] in, This is a simplified form of joint torque. and They are respectively and abbreviated form, The inertia matrix, This is the matrix of Coriolis force and centrifugal force. The gravitational torque vector, For Jacobian matrices, External force / torque generated by the load at the actuator end.

[0032] S20, based on data acquired from multiple sensors, utilizes an improved RRT The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection.

[0033] For example, standard RRT Uniform random sampling results in slow convergence. The dynamic bias probability described in this embodiment is expressed as:

[0034] in, The current iteration number The dynamic bias probability mentioned below, This is the initial bias probability. For the maximum bias probability, Let be the time constant. At each sampling, with probability... Directly using the target configuration As sampling points, with probability Uniform random sampling is performed in the joint space. Initially, the bias probability is low to ensure exploration coverage, and later the bias probability gradually increases to accelerate convergence to the target.

[0035] Standard RRT The algorithm uses Euclidean distance in joint space to measure the equal weight of each joint dimension:

[0036] Among them, the standard RRT The algorithm uses the distance between joints a and b. and These are the configurations of joints a and b, respectively.

[0037] However, the contributions of different joints to the end effector's Cartesian spatial pose change vary greatly. For example, a small rotation of the lumbar joint (joint 1) may lead to a large translation of the actuator, while a large rotation of the wrist joints (joints 5 and 6) may only cause a small posture adjustment of the actuator. Therefore, in the embodiments of this application, the weighted distance between joints a and b is expressed as:

[0038] in, The weighted distance between joints a and b. and These are the configurations of joints a and b, respectively. The superscript T indicates transpose. The weight matrix is ​​represented as follows:

[0039]

[0040] in, Let be the weight of the i-th joint, and n be the total number of joints. is the i-th column of the Jacobian matrix (corresponding to the contribution of joint i to the end velocity). This represents the intermediate configuration for joints a and b. This weighted metric allows joints that significantly influence the pose of the execution end to achieve higher resolution in distance calculations, thereby making the expansion of the search tree in Cartesian space more uniform.

[0041] In collision detection, an octree map (5mm resolution) is constructed from the 3D point cloud obtained by fusing data from multiple sensors, reducing the query complexity of collision detection from brute-force search. Down to , where N p The number of points in a 3D point cloud, such that RRT The bottleneck of collision detection is effectively alleviated. At the same time, by approximating each link in the robotic arm as a sphere chain model, the complex geometric collision detection is simplified to a distance query between the sphere and the octree node, further improving the detection efficiency.

[0042] S30, the collision-free path is smoothed using a 5th-order B-spline curve to obtain a smooth path.

[0043] For example, an improved RRT is used. The path point sequence obtained by the algorithm search Typically, this involves sharp corners, where M represents the number of path points. Direct execution would cause the robotic arm to experience sudden speed changes and mechanical impacts at the corners. This application adopts... Rank ( The path points are smoothed using a uniform B-spline curve to generate a smoothing effect in the joint space. Continuous trajectory curve:

[0044] in, This is the smoothed path after smoothing. for B-order spline basis functions, For the parameter variables of the curve, Let r be the r-th control point, where r is the index of the control point. The control points are determined by minimizing the following objective function:

[0045] in, For the first Path points In the parameter domain of B-spline curves The corresponding parameter value assigned above, where m is the index of the path point. To improve RRT The algorithm search output of the first Each path point is a complete joint configuration vector of the robotic arm at that path point (containing all...). (angle values ​​of each joint), first item The second term ensures that the smooth trajectory passes through (or approximately passes through) the original path points. For regularization terms ( ), restricting the second derivative of the curve to avoid excessive oscillation, for The second derivative of . B-splines ensure the continuity of the trajectory in terms of position, velocity, and acceleration. (Continuous), eliminating abrupt acceleration changes at corners.

[0046] S40. Based on the smooth path, a dual objective function based on total motion time and total energy consumption is established. The weighted Chebyshev method is used to transform the dual objective function into a single objective problem. Under the constraints of the grasping configuration, the sequential quadratic programming method is used to solve the problem and obtain the optimal path.

[0047] For example, the biobjective function based on total motion time and total energy consumption is established, including: S41, Establish the mapping relationship between the parameters of the 5th order B-spline curve and time, and then determine the position, velocity and acceleration of the joint respectively; S42, the time is discretized into L segments to determine the total motion time; S43, the total energy consumption is expressed as follows:

[0048] in, This represents the total energy consumption. l Here, n is the index of the time segment, n is the total number of joints, and i is the index of the joint. and For two adjacent moments, This represents the torque at the i-th joint, which is determined based on the joint's position, velocity, and acceleration. Let be the angular velocity of the i-th joint. This indicates taking the absolute value because the motor consumes electrical energy regardless of whether it is in driving mode (power is positive) or braking mode (power is negative) (braking energy is dissipated through the braking resistor). Joint torque. Based on the inverse dynamics model , , and load quality Calculation. In the numerical implementation, the above integral is calculated by discretization using the fourth-order Runge-Kutta method.

[0049] Specifically, in S41, the parameters of the 5th-order B-spline curve are... With time The mapping relationship is represented as ,in It is a monotonically increasing time-parameterized function. This represents the total motion time. Then the joint position... ,speed and acceleration They are respectively:

[0050] in, and For about The first and second derivatives, and They are respectively The first and second derivatives.

[0051] The optimization variable is the discrete representation of the time-parameterized function s: Evenly divided into Segments, with a time interval between each segment. ( Then the total motion time Total exercise time It can also be expressed as:

[0052] because and There is usually a contradiction: shortening the time requires increasing acceleration, which in turn increases energy consumption. In this application, the weighted Chebyshev method is used to transform the biobjective function into a single objective problem.

[0053] in, and These are the ideal optimal values ​​for each of the two objectives, i.e., the optimal values ​​when optimized individually. and The Nadir point value represents the worst-case value of the objective when optimizing another objective. The denominator term serves to normalize the two objectives, ensuring that their dimensions are consistent. and For user-configurable weighting coefficients, In scenarios where loading efficiency is the priority, the following settings are implemented: In scenarios where energy conservation is the priority, .

[0054] In solving the single-objective problem, the constraints include all the aforementioned grasping configuration constraints. Sequential Quadratic Programming (SQP) is used, with initial solutions allocated at equal time intervals. In each iteration of SQP, the original problem is approximated twice at the current solution, solving the quadratic programming subproblems to obtain the search direction, and then determining the step size through line search. The convergence condition is that the residual norm of the KKT conditions (Kuhn-Karush-Tucker conditions) is less than... Or it can reach the maximum number of iterations of 200.

[0055] S50, during the execution of the optimal path, an octree update algorithm is used to maintain the dynamic environment map, and when a new obstacle is detected in the safe corridor, local replanning is performed; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

[0056] For example, industrial loading and unloading sites are not entirely static, as conveyor belts continuously transport new bags of material, and forklifts or personnel may temporarily enter the robotic arm's workspace. Real-time data from multiple sensors is read at a frequency of at least 50Hz, and a dynamic environment map is maintained using an incremental octree update algorithm. Newly added point clouds are rasterized and inserted into the octree, while the grids corresponding to disappeared objects are marked as free space using ray casting.

[0057] The core idea of ​​the incremental octree update algorithm is: instead of rebuilding the complete octree map in each frame, it locally corrects the existing map based on the new observation data of the current frame, and only updates the nodes corresponding to the spatial regions that have changed, thereby keeping the computational load at a level proportional to the amount of environmental change.

[0058] Specifically, each leaf node in an octree Maintain a log-odds occupancy value This is used to characterize the probability that the grid is occupied by an obstacle. When a new frame of point cloud data arrives, ray casting is performed on each measurement point from the sensor origin: the grid node where the ray ends is updated with an increased occupancy probability. The occupancy probability of the grid nodes along the ray traversal path is reduced during the update. ,in and All are probability correction values, and , At the same time, upper and lower limit clamps are applied to the logarithmic probability value to prevent overconfidence caused by long-term accumulation and to ensure that the node state can respond in a timely manner when the environment changes.

[0059] After the update is complete, nodes are classified into three states: "occupied," "free," or "unknown" based on logarithmic probability values. When the collision detection module queries, nodes in the "occupied" and "unknown" states are considered potential obstacles, and the robotic arm is only allowed to pass through areas in the "free" state to ensure safe movement. Through this incremental update mechanism, each frame only processes the local nodes involved in the current observation, and the map update time per frame is controlled within 5ms, meeting the timeliness requirements of real-time dynamic obstacle avoidance.

[0060] During the optimal path execution process, a configuration checkpoint is extracted every 10ms along the path to measure the expansion of the robot arm envelope under that configuration. This is a safe corridor. When real-time sensing detects a new obstacle within the safe corridor, local trajectory replanning is triggered.

[0061] The strategy for local replanning is to use the current state of the robotic arm joints. Starting from the original target, capture the configuration. The endpoint is only the local section of the security corridor that has been breached. Re-execute RRT Searching and B-spline smoothing generate alternative trajectory segments. The alternative segments and the original trajectory meet at the splicing point. and The following conditions must be met for continuity of position, velocity, and acceleration:

[0062]

[0063] in, The alternative trajectory segments generated for local replanning are at the same splicing point. The joint configuration at that location refers to the angle values ​​that the robotic arm's joints are required to achieve at that position for the new trajectory. At the splicing point, the original planned trajectory (i.e., the trajectory that was generated and being executed before the security corridor was intruded upon) is... The joint configuration at that location refers to the angle values ​​that each joint of the robotic arm should originally reach at that position. and They are respectively The first and second derivatives, and They are respectively The first and second derivatives. The same applies to the other parameters.

[0064] These six boundary conditions (three at each splice point) can be satisfied by the control point degrees of freedom of a 5th-order B-spline. The continuity constraint ensures a smooth transition of the robotic arm's motion at switching moments, avoiding mechanical vibration and bagged material slippage caused by sudden stops or abrupt changes.

[0065] When an obstacle is detected to suddenly appear at extremely close range in the current configuration (distance less than the emergency threshold) If the computation time for local replanning is insufficient to complete before a collision, the system triggers emergency braking. Emergency braking does not employ a sudden stop; instead, it performs a smooth deceleration to zero according to the maximum permissible deceleration of each joint.

[0066] in, This is the time required for deceleration. Environmental perception is continuously updated during deceleration, and once an obstacle leaves the safe zone, a complete path replanning from zero speed to the target configuration is immediately triggered.

[0067] Compared with the prior art, the method proposed in this application has the following advantages: 1. Gripping Configuration Constraint Modeling. Existing technologies do not consider the process constraints of flexible gripping of bagged materials, which easily leads to bag slippage, collisions, joint overload, and low gripping success rate. The embodiments of this application transform process requirements into mathematical constraints, ensuring gripping accuracy and robotic arm operation safety from the source, maximizing the effective component of gripping force, and providing a unified constraint benchmark for subsequent path planning.

[0068] 2. Improve RRT Pathfinding. Existing standard RRT Slow convergence speed, unreasonable distance measurement, and low collision detection efficiency result in search times that cannot meet the real-time requirements of industrial applications. The embodiments of this application improve convergence speed by 65.6%, reduce single path search time from 280ms to 95ms, shorten path length by approximately 18%, and achieve more uniform path expansion in Cartesian space. Collision detection efficiency is significantly improved, fully meeting the real-time requirements of industrial environments.

[0069] 3. The 5th-order B-spline trajectory is smooth. Existing discrete path points have sharp corners, which, when directly executed, cause sudden changes in velocity / acceleration and mechanical shock, affecting equipment lifespan and gripping stability. The embodiment of this application generates C... 4 Continuous trajectory eliminates sudden acceleration changes at inflection points, reducing mechanical impact by 38% to 52%, extending the service life of robotic arms and reducers, and improving gripping stability.

[0070] 4. Dual-objective trajectory optimization of time and energy consumption. Existing technologies only optimize a single objective, failing to balance time efficiency and energy consumption, resulting in high operational energy consumption and poor adaptability. The embodiments of this application achieve Pareto optimality for both time and energy consumption, supporting dual-mode configuration of efficiency and energy saving; the efficiency-first mode saves 13.4% of time and reduces energy consumption by 10.1%, while the energy-first mode reduces energy consumption by 24.6%, catering to the needs of different industrial scenarios and reducing operating costs.

[0071] 5. Dynamic obstacle avoidance and real-time trajectory correction. Existing technologies lack dynamic environment perception and trajectory correction capabilities, making them prone to collisions in dynamic scenes and exhibiting poor operational robustness. This application's embodiment utilizes a 50Hz high-frequency sensor to perceive the dynamic environment, monitors safety corridors to achieve rapid local replanning, and performs emergency smooth braking to avoid collisions; trajectory stitching meets continuity requirements, supporting fully automated unmanned operation in dynamic environments, significantly improving operational robustness.

[0072] This application also provides a time-energy dual-objective optimized path planning system for bagged material grasping, such as... Figure 2 As shown, the system includes the following modules: The grab configuration constraint construction module is used to construct grab configuration constraint conditions, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints. Improved RRT The path search module is used to search for data from multiple sensors, utilizing an improved RRT. The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection. A 5th-order B-spline trajectory smoothing module is used to smooth the collision-free path using a 5th-order B-spline curve to obtain a smooth path. The time-energy dual-objective optimization module is used to establish a dual-objective function based on the total motion time and total energy consumption on the basis of the smooth path. The weighted Chebyshev method is used to transform the dual-objective function into a single-objective problem. Under the constraints of the grasping configuration, the sequential quadratic programming method is used to solve the problem and obtain the optimal path. The dynamic obstacle avoidance and trajectory correction module is used to maintain a dynamic environment map using an octree update algorithm during the execution of the optimal path, and to perform local replanning when a new obstacle is detected in the safe corridor; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

[0073] This application also provides a computer storage medium storing a plurality of computer instructions for causing a computer to execute the above-described method.

[0074] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0075] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A time-energy dual-objective optimization path planning method for bagged material grasping, characterized in that, include: S10, construct the gripping configuration constraints, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints; S20, based on data acquired from multiple sensors, utilizes an improved RRT The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection. S30, the collision-free path is smoothed using a 5th-order B-spline curve to obtain a smooth path; S40, based on the smooth path, establish a dual objective function based on total motion time and total energy consumption, and use the weighted Chebyshev method to convert the dual objective function into a single objective problem. Under the constraints of the grasping configuration, use the sequential quadratic programming method to solve the problem and obtain the optimal path. S50, during the execution of the optimal path, an octree update algorithm is used to maintain the dynamic environment map, and when a new obstacle is detected in the safe corridor, local replanning is performed; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

2. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, The approach direction constraint is that the angle between the actuator and the approach direction and the normal vector of the principal plane of the bag body is less than or equal to the maximum approach deviation angle; the collision avoidance constraint is that the minimum Euclidean distance between the space occupied by the robotic arm link and the actuator assembly and the space occupied by any obstacle is greater than or equal to the safety margin; the joint kinematic constraint is that the angle, angular velocity and angular acceleration of each joint are within the hardware allowable range; the joint torque constraint is that the torque of each joint is less than or equal to the maximum torque that can be withstood.

3. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 2, characterized in that, In the joint moment constraint, the joint moment is calculated by the inverse dynamics model of the robotic arm, which adopts the Newton-Euler recursive formula.

4. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, The dynamic bias probability is expressed as: in, The current iteration number The dynamic bias probability mentioned below, This is the initial bias probability. For the maximum bias probability, is the time constant.

5. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, The weighted distance between joints a and b is expressed as: in, The weighted distance between joints a and b. and These are the configurations of joints a and b, respectively. The superscript T indicates transpose. The weight matrix is ​​represented as follows: in, Let be the weight of the i-th joint, and n be the total number of joints. Let i be the i-th column of the Jacobian matrix. This is the intermediate configuration between joints a and b. This represents the L2 norm.

6. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, The improved RRT In collision detection, the algorithm approximates each link of the robotic arm as a ball chain model.

7. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, Establish the biobjective function based on total motion time and total energy consumption, including: Establish the mapping relationship between the parameters of the 5th order B-spline curve and time, and then determine the position, velocity and acceleration of the joint respectively; The time is discretized into L segments to determine the total motion time; The total energy consumption is expressed as: in, The total energy consumption is represented by l, where l is the index of the time period, n is the total number of joints, and i is the index of the joint. and For two adjacent moments, This represents the torque at the i-th joint, which is determined based on the joint's position, velocity, and acceleration. Let be the angular velocity of the i-th joint. This indicates taking the absolute value.

8. The time-energy dual-objective optimized path planning method for bagged material grasping according to claim 1, characterized in that, The local replanning is as follows: based on the current joint state... Starting from the original target, capture the configuration. The endpoint is only the local section of the security corridor that is intruded upon. Re-execute RRT Search and B-spline smoothing to generate alternative trajectory segments.

9. A system applying the time-energy dual-objective optimization path planning method for bagged material grasping as described in any one of claims 1-8, characterized in that, include: The grab configuration constraint construction module is used to construct grab configuration constraint conditions, which include approach direction constraints, collision avoidance constraints, joint kinematic constraints, and joint moment constraints. Improved RRT The path search module is used to search for data from multiple sensors, utilizing an improved RRT. The algorithm searches for collision-free paths in joint space, and the improved RRT... The algorithm samples the joint space based on a dynamic bias probability during the search process, and this dynamic bias probability increases with the number of iterations; the improved RRT is used during the search process. The algorithm uses the Jacobian matrix to calculate the weighted distance of each joint, enabling joints that have a significant impact on the pose of the execution end to obtain higher resolution in the weighted distance calculation; simultaneously, the improved RRT The algorithm constructs an octree map from the 3D point cloud collected by multiple sensors during the collision detection process, reducing the query complexity of collision detection. A 5th-order B-spline trajectory smoothing module is used to smooth the collision-free path using a 5th-order B-spline curve to obtain a smooth path. The time-energy dual-objective optimization module is used to establish a dual-objective function based on the total motion time and total energy consumption on the basis of the smooth path. The weighted Chebyshev method is used to transform the dual-objective function into a single-objective problem. Under the constraints of the grasping configuration, the sequential quadratic programming method is used to solve the problem and obtain the optimal path. The dynamic obstacle avoidance and trajectory correction module is used to maintain a dynamic environment map using an octree update algorithm during the execution of the optimal path, and to perform local replanning when a new obstacle is detected in the safe corridor; if the computation time of local replanning is insufficient to complete before the collision, emergency braking is triggered to reduce the speed of each joint to zero.

10. A computer storage medium, characterized in that, The computer storage medium stores a plurality of computer instructions, which are used to cause the computer to perform the method described in any one of claims 1-8.

Images