Environment perception based robot obstacle avoidance trajectory planning method and system

Abstract

The application relates to an environment perception-based mechanical arm obstacle avoidance trajectory planning method and system, which comprises the following steps: acquiring joint angle data and end effector position data of a mechanical arm, establishing a joint space for mechanical arm work according to the joint angle data and the end effector position data; acquiring environment point cloud data, extracting obstacle information based on the environment point cloud data, and acquiring a safe operation space of the mechanical arm through the obstacle information; acquiring a current joint angle combination of the mechanical arm, converting a position difference between the current joint angle combination and a target position into an angle increment in the joint space, and adjusting an angle value exceeding the boundary of the safe operation space, so as to complete obstacle avoidance trajectory planning of the mechanical arm. The application can realize a grabbing task while avoiding obstacles.

Description

A method and system for obstacle avoidance trajectory planning for robotic arms based on environmental perception Technical Field

[0001] This invention relates to the field of robotics, and in particular to a method and system for planning obstacle avoidance trajectories for a robotic arm based on environmental perception. Background Technology

[0002] Robotics technology plays an irreplaceable role in industrial automation, medical assistance, and service fields. In particular, the precise control capability of robotic arms directly determines the overall performance and application scope of automation systems. As application scenarios become increasingly complex, traditional robotic arm control methods exhibit significant limitations when facing dynamic environments and complex tasks. Currently, mainstream robotic arm control methods primarily rely on pre-defined trajectory planning and fixed control parameters. While this approach performs well in structured environments, it often fails to effectively cope with changes in the environment or the appearance of unexpected obstacles, leading to task failure or safety hazards.

[0003] The core challenge in robotic arm control lies in achieving an effective conversion between joint space and Cartesian space. The physical structure of a robotic arm dictates that its basic motion is achieved through changes in joint angles, while practical tasks typically require precise positioning of the end effector in a Cartesian coordinate system. This difference in spatial representation makes precise control directly in Cartesian space difficult, especially when multiple constraints need to be considered simultaneously. This control challenge further leads to the perception and decision-making problems of robotic arms in complex environments. The system needs to process a large amount of sensor information in real time and make optimal decisions, including multi-dimensional information such as obstacle distance, target object position, and end effector state. The comprehensive processing of this information places extremely high demands on traditional control algorithms.

[0004] Therefore, designing a control system that can seamlessly convert the joint control of a robotic arm into Cartesian spatial position control, while comprehensively considering environmental state information and making intelligent decisions, has become a key issue in improving the ability of robotic arms to perform grasping tasks in complex environments. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide a robotic arm obstacle avoidance trajectory planning method based on environmental perception. This invention can dynamically plan a reasonable obstacle avoidance trajectory, enabling the robotic arm to react quickly and effectively avoid complex and ever-changing environments and dynamically appearing obstacles, thereby improving its adaptability in complex environments and task success rate.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A method for planning obstacle avoidance trajectory for a robotic arm based on environmental perception, comprising:

[0008] Acquire the joint angle data and end effector position data of the robotic arm, and establish the joint space for the robotic arm's operation based on the joint angle data and end effector position data;

[0009] Acquire environmental point cloud data, extract obstacle information based on the environmental point cloud data, and obtain the safe operating space of the robotic arm through the obstacle information;

[0010] The current joint angle combination of the robotic arm is obtained, the positional difference between the current joint angle combination and the target position is converted into an angle increment in the joint space, and the angle value exceeding the boundary of the safe operating space is adjusted to complete the obstacle avoidance trajectory planning of the robotic arm.

[0011] Optionally, establishing the joint space includes: converting the end effector position data into pose parameters in a Cartesian coordinate system, and combining the joint angle data to establish the joint space in which the robotic arm operates.

[0012] Optionally, obtaining the safe operating space of the robotic arm includes:

[0013] A multi-target sensor system, consisting of a depth camera and a lidar, is used to collect environmental point cloud data. Obstacle information is extracted based on the environmental point cloud data. A transformation matrix between the multi-target sensor and the robotic arm is established according to the relationship between the position of the multi-target sensor and the robotic arm base. The obstacle information is spatially registered using the transformation matrix to generate a unified 3D scene model containing the robotic arm and environmental objects. The minimum distance between each joint of the robotic arm and the environmental objects is calculated using the unified 3D scene model. When the minimum distance is less than a safety threshold, it is marked as a potential collision risk area. When the minimum distance is greater than the safety threshold, it is marked as a safe operating space.

[0014] Optionally, converting the positional difference between the current joint angle combination of the robotic arm and the target position into an angle increment in the joint space includes:

[0015] The current joint angle combination is mapped to the end effector position in Cartesian space by using a forward kinematics model. The Euclidean distance between the current position and the target position of the end effector is obtained. The Jacobian matrix is ​​constructed using the partial derivatives of the forward kinematics model. The singular points of the Jacobian matrix are processed to obtain a stable inverse Jacobian matrix. The Euclidean distance is converted into the angle increment in joint space using the inverse stable inverse Jacobian matrix.

[0016] Optionally, mapping the current combination of joint angles to the end effector position in Cartesian space using the forward kinematics model includes:

[0017] p = f(q)

[0018] Where q is the joint angle vector, and p = [x, y, z] T Let T be the position coordinates of the end effector, T be the transpose, and f be the forward kinematic function, which is usually a nonlinear function that describes the relationship between the joint angle and the end effector position.

[0019] Optionally, adjusting the angle value beyond the boundary of the safe operating space includes:

[0020]

[0021] Where, P n It is the angle adjustment value of the joint position of the robotic arm obtained after the nth correction, r represents the maximum radius from the geometric center of the obstacle to its surface, γ is the correction coefficient, A is the minimum distance between the joint position of the robotic arm and the geometric center of the obstacle, and B is the direction vector of the Euclidean distance.

[0022] To achieve the above objectives, the present invention also provides an environment-aware robotic arm obstacle avoidance trajectory planning system, comprising:

[0023] The joint space construction module is used to acquire the joint angle data and end effector position data of the robotic arm, and to establish the joint space for the robotic arm to work based on the joint angle data and end effector position data.

[0024] The operating space extraction module is used to acquire environmental point cloud data, extract obstacle information based on the environmental point cloud data, and obtain the safe operating space of the robotic arm through the obstacle information.

[0025] The obstacle avoidance trajectory planning module is used to obtain the current joint angle combination of the robotic arm, convert the positional difference between the current joint angle combination and the target position into the angle increment in the joint space, and adjust the angle value that exceeds the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm.

[0026] Optionally, the joint space construction module includes:

[0027] The data acquisition unit is used to acquire the joint angle data of the robotic arm and the position data of the end effector;

[0028] The joint space construction unit is used to convert the end effector position data into pose parameters in the Cartesian coordinate system, and combine it with joint angle data to establish the joint space for the operation of the robotic arm.

[0029] Optionally, the operation space extraction module includes:

[0030] An obstacle information extraction unit is used to collect environmental point cloud data using a multi-target sensor consisting of a depth camera and a lidar, and to extract obstacle information based on the environmental point cloud data.

[0031] The operation space extraction unit is used to establish a transformation matrix between the multi-target sensors and the robotic arm base based on the position of the multi-target sensors and the relationship between the multi-target sensors and the robotic arm base. The transformation matrix is ​​used to spatially register the obstacle information to generate a unified three-dimensional scene model containing the robotic arm and environmental objects. The unified three-dimensional scene model is used to calculate the minimum distance between each joint of the robotic arm and the environmental objects. When the minimum distance is less than a safety threshold, it is marked as a potential collision risk area. When the minimum distance is greater than the safety threshold, it is marked as a safe operation space.

[0032] Optionally, the obstacle avoidance trajectory planning module includes:

[0033] The first obstacle avoidance trajectory planning unit is used to map the current joint angle combination to the end effector position in Cartesian space through the forward kinematics model, obtain the Euclidean distance between the current position of the end effector and the target position, construct the Jacobian matrix using the partial derivatives of the forward kinematics model, process the singular points of the Jacobian matrix to obtain a stable inverse Jacobian matrix, and use the inverse stable inverse Jacobian matrix to convert the Euclidean distance into the angle increment in joint space.

[0034] The second obstacle avoidance trajectory planning unit is used to adjust the angle value that exceeds the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm.

[0035] The beneficial effects of this invention are as follows:

[0036] This invention acquires environmental point cloud data through a multi-target sensor consisting of a depth camera and a lidar, enabling precise perception of obstacle information in the surrounding environment, including the location, shape, and size of obstacles. Furthermore, the established unified 3D scene model allows the robotic arm to understand its spatial relationship with environmental objects in real time and accurately, thereby dynamically planning a reasonable obstacle avoidance trajectory. This enables the robotic arm to react quickly and effectively avoid complex and ever-changing environments and dynamically appearing obstacles, improving its adaptability in complex environments and task success rate.

[0037] This invention maps joint angle combinations to the Cartesian space position of the end effector using a forward kinematics model and transforms them using the Jacobian matrix and its inverse matrix. This solves the problem of the robotic arm's transformation between joint space and Cartesian space, enabling the robotic arm to operate more flexibly in both spaces when planning its trajectory. At the same time, it ensures the accuracy of the end effector's position control and improves the robotic arm's ability to complete complex tasks.

[0038] In the process of converting positional differences into joint space angle increments, this invention processes the singular points of the Jacobian matrix to obtain a stable inverse Jacobian matrix, thereby avoiding motion singularities or instability in trajectory planning. This ensures the stability and reliability of the robotic arm throughout the entire motion process, making the planned trajectory more optimized and reasonable, reducing jitter and errors during the robotic arm's movement, and improving motion accuracy.

[0039] This invention calculates the minimum distance between each joint of the robotic arm and environmental objects, and compares it with a safety threshold to clearly delineate potential collision risk areas and safe operating spaces. During movement, the robotic arm can always plan obstacle avoidance trajectories within the safe operating space, avoiding entering potential collision risk areas. This effectively reduces the probability of collisions between the robotic arm and environmental objects, ensuring the safety of the robotic arm itself and surrounding equipment. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 is a flowchart of a robotic arm obstacle avoidance trajectory planning method based on environmental perception according to an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] As shown in Figure 1, this embodiment discloses a robotic arm obstacle avoidance trajectory planning method based on environmental perception, including: acquiring joint angle data and end effector position data of the robotic arm; establishing the joint space for the robotic arm's operation based on the joint angle data and end effector position data; acquiring environmental point cloud data; extracting obstacle information based on the environmental point cloud data; obtaining the safe operating space of the robotic arm through the obstacle information; acquiring the current joint angle combination of the robotic arm; converting the positional difference between the current joint angle combination and the target position into angle increments in the joint space; and adjusting the angle values ​​exceeding the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm. Optionally, establishing the joint space includes: converting the end effector position data into pose parameters in a Cartesian coordinate system and combining it with the joint angle data to establish the joint space for the robotic arm's operation.

[0045] Specifically, encoder data from each joint of the robotic arm is collected, converted into standard angle values, and the joint angle sequence and timestamp are recorded to construct a joint space state vector, obtaining a real-time data stream of the robotic arm joint configuration. End-effector position sensor data is acquired and converted into position and attitude parameters in a Cartesian coordinate system. Combined with joint angle data, a complete pose description of the end-effector in three-dimensional space is established. A geometric model of the robotic arm's workspace is constructed. Joint angle combinations are randomly generated in the joint space using Monte Carlo sampling, mapped to Cartesian space, and the workspace boundary is represented by point cloud data. Further analysis of the robotic arm's motion constraints, including joint angle limitations, velocity limitations, and acceleration limitations, generates a set of motion constraint parameters. The K-means clustering algorithm is applied to divide the workspace into regions. Based on the ease of access for the end-effector, singularity regions and highly operable regions in the workspace are identified, and the operational characteristics of different regions are marked. The joint angles, end-effector position, workspace boundary, and motion constraint parameters are integrated to establish a robotic arm state database. A data index structure enables fast querying and state matching, supporting subsequent path planning and task execution.

[0046] For example, the joint encoder of a six-axis industrial robotic arm may output raw values ​​such as [3276,16384,8192,-4096,12288,6144], which are then converted into standard angle values ​​[-45°, 30°, 15°, -7.5°, 22.5°, 11.25°] through linear mapping and appended with a timestamp "2023-05-15 14:30:25.342" to form a joint space state vector.

[0047] Specifically, end effector position sensing can be achieved through a position sensor installed at the end.

[0048] For example, the raw data [245,178,320,0.707,0,0,0.707] output by the sensor can be transformed to obtain the position [245mm,178mm,320mm] and attitude (quaternion representation) [0.707,0,0,0.707] in the Cartesian coordinate system. This is equivalent to the end effector being located at a specific position in the workspace and rotating 90° around the Z-axis.

[0049] In one embodiment, workspace modeling can be achieved through Monte Carlo sampling. By generating 10,000 random joint angle combinations in joint space, such as θ1∈[-180°, 180°], θ2∈[-45°, 90°], etc., and then mapping them to Cartesian space using forward kinematic equations, point cloud data is formed. This point cloud data is processed using a convex hull algorithm to obtain the geometric boundaries of the robotic arm's workspace, facilitating an intuitive understanding of the robotic arm's working range.

[0050] It should be noted that motion constraint analysis is crucial for safe operation. For example, joint angle constraints might be set as follows: first joint [-180°, 180°], second joint [-45°, 90°]; velocity constraints as [120° / s, 90° / s, 150° / s, 180° / s, 180° / s, 210° / s]; and acceleration constraints as [180° / s²]. 2 135° / s 2 225° / s 2 270° / s 2 270° / s 2 315° / s 2 By combining obstacle information in the working environment, such as the workbench position [500mm, 0mm, -100mm], a complete set of motion constraint parameters can be generated.

[0051] Applying the K-means clustering algorithm (k=5) to divide the workspace into regions can identify singularity regions (such as when joints are extended or aligned) and highly operable regions. For example, the region near coordinates [300mm, 0mm, 400mm] may be marked as a highly operable region with an operability index of 0.85, while the region near [500mm, 0mm, 100mm] may be marked as a near-singular region with an operability index of only 0.12.

[0052] Furthermore, obtaining the safe operating space of the robotic arm includes: collecting environmental point cloud data using a multi-target sensor consisting of a depth camera and a LiDAR; extracting obstacle information based on the environmental point cloud data; establishing a transformation matrix between the multi-target sensor and the robotic arm based on the position of the multi-target sensor and the relationship between the multi-target sensor and the robotic arm base; using the transformation matrix to spatially register the obstacle information; generating a unified 3D scene model containing the robotic arm and environmental objects; and calculating the minimum distance between each joint of the robotic arm and environmental objects using the unified 3D scene model. When the minimum distance is less than a safety threshold, it is marked as a potential collision risk area; when the minimum distance is greater than the safety threshold, it is marked as a safe operating space.

[0053] Specifically, a multi-target sensor system composed of a depth camera and LiDAR is used to collect environmental point cloud data from different angles, recording the 3D information of obstacles and the pose data of target objects. Simultaneously, data from the robotic arm joint encoders is read to obtain the current state data of the robotic arm. The collected raw point cloud data is processed using a voxel filtering algorithm for noise reduction and sparsification, removing outliers and redundant information to improve subsequent processing efficiency and obtain clear point cloud data with appropriate information density. Geometric features, including key features such as planes, edges, and curved surfaces, are extracted from the filtered point cloud data. The RANSAC algorithm is used to identify the main structural elements in the environment, marking object boundaries and surface normal vector information. Based on the sensor installation position and the relationship between the sensor and the robotic arm base, a transformation matrix between the sensor and the robotic arm is established. An iterative nearest-point algorithm is used to spatially register data from different sensors, achieving a unified transformation of all environmental data to the robotic arm coordinate system. An octree spatial partitioning structure is constructed using the transformed point cloud data to represent the environment in voxel form, marking occupied space and free space, generating a unified 3D scene model containing the robotic arm and environmental objects. Based on the constructed scene model, the minimum distance between each joint of the robotic arm and environmental objects is calculated. When the distance is less than the safety threshold, it is marked as a potential collision risk area, and the area with a distance greater than the safety threshold is marked as a safe operating space.

[0054] For example, in a robotic arm environmental perception system, multi-sensor fusion is key to improving perception accuracy. A six-axis robotic arm workstation can be equipped with an Intel RealSense D435i depth camera and a Velodyne VLP-16 LiDAR simultaneously, acquiring environmental data from above and the side of the worktable respectively, forming complementary perception perspectives. Voxel filtering is an important step in point cloud data preprocessing.

[0055] Specifically, the original point cloud can be divided into a cubic grid with a side length of 0.01 meters. Points within each grid are merged into a single point. This reduces the original scene containing approximately 2 million points to approximately 50,000 points, while preserving key structural information and significantly improving subsequent processing speed. This is relevant in the feature extraction stage.

[0056] It should be noted that the RANSAC algorithm can effectively identify key geometric elements in the environment. For example, in an assembly workstation, the algorithm can extract the worktable plane (the largest planar feature), the cylindrical surface features of the part to be grasped (with a radius of approximately 0.03 meters), and the edge features of the assembly base from the point cloud, providing crucial reference information for subsequent operations. A robust sensor and robotic arm coordinate system is a prerequisite for achieving precise operation.

[0057] In one possible implementation, a calibration plate can be placed in the robotic arm's workspace and simultaneously observed by a depth camera and LiDAR. The transformation matrices T1 and T2 from the two sensors to the calibration plate can be calculated. Then, using the known transformation matrix T3 from the robotic arm base to the calibration plate, the transformation relationship from the sensor to the robotic arm base can be obtained: Tsensor to base = T3·(T1)-1 or T3·(T2)-1. Octree spatial partitioning can efficiently represent the 3D environment. For example, dividing the workspace into an octree structure with a minimum resolution of 0.02 meters can form an environment representation containing approximately 100,000 voxels for a 2m × 2m × 2m workspace. Each voxel is labeled with one of three states: occupied, free, or unknown, achieving a compact representation of the environment. Collision risk assessment is essential for safe operation.

[0058] In one embodiment, when the distance between the third joint of the robotic arm and the edge of the workbench is detected to be less than 0.1 meters, the system will mark the area as a yellow warning zone. If the distance is less than 0.05 meters, it will be marked as a red danger zone, and the movement speed of the robotic arm will be adjusted accordingly or obstacle avoidance planning will be triggered.

[0059] Furthermore, converting the positional difference between the current joint angle combination of the robotic arm and the target position into angle increments in joint space includes: mapping the current joint angle combination to the end effector position in Cartesian space using a forward kinematics model, obtaining the Euclidean distance between the current position of the end effector and the target position, constructing a Jacobian matrix using the partial derivatives of the forward kinematics model, handling the singularities of the Jacobian matrix to obtain a stable inverse Jacobian matrix, and using the inversely stable inverse Jacobian matrix to convert the Euclidean distance into angle increments in joint space.

[0060] Specifically, unified scene model data and safe operating space boundaries are acquired, and obstacles in the scene are represented in three-dimensional space to obtain the precise boundaries of the robotic arm's movable range. The positional difference between the current joint angle state and the target pose is calculated. The joint angle combination in Cartesian space is mapped to the end effector position using forward kinematics equations, and the Euclidean distance between the current position and the target position is determined. A Jacobian matrix is ​​constructed, and the sensitivity matrix of the impact of joint angle changes on the end effector position is calculated based on the robotic arm's geometric parameters. Singular value decomposition is used to process the singular points of the Jacobian matrix, resulting in a stable inverse Jacobian matrix. An inverse kinematics algorithm is applied, using the inverse Jacobian matrix to convert position increments in Cartesian space into angle increments in joint space. Joint limit checks are performed on the conversion results, and angle values ​​exceeding mechanical limits and safe operating space boundaries are adjusted.

[0061] When a robotic arm needs to grasp parts on an industrial production line, the workbench, surrounding equipment, and other obstacles must first be represented as three-dimensional geometric shapes. For example, a conveyor belt can be represented as a cuboid 20 cm high, 50 cm wide, and 200 cm long, and a fixed support can be represented as a cylinder 5 cm in diameter and 150 cm high. These geometric representations form the constraints on the robotic arm's movement, clearly defining the safe range of its movement.

[0062] In one possible implementation, when the robotic arm needs to move from its current position (joint angles [30°, 45°, -20°, 60°, 10°, 0°]) to a target position (end-effector coordinates [500mm, 300mm, 200mm]), the system first calculates the position difference. Using the forward kinematics equations, the current joint angles are mapped to the end-effector position in Cartesian space [450mm, 250mm, 180mm], resulting in a position difference of [50mm, 50mm, 20mm], with an Euclidean distance of approximately 73.5mm.

[0063] Specifically, the construction of the Jacobian matrix is ​​crucial for motion planning. For a 6-DOF robotic arm, its Jacobian matrix is ​​a 6×6 matrix, representing the influence of the angular velocity of each joint on the linear and angular velocities of the end effector. When the robotic arm approaches a singularity point (such as when the arm is fully extended), singularity issues are handled through singular value decomposition and a singular value threshold (such as 0.01) to ensure stable calculation of the inverse Jacobian matrix.

[0064] In one embodiment, inverse kinematics calculations convert position increments [50mm, 50mm, 20mm] in Cartesian space into angle increments [5°, 8°, -3°, 7°, 2°, 1°] in joint space. The system then checks joint limits; for example, if the angle of the third joint changes to -23°, close to its lower limit of -25°, adjustments are needed to allow for a safety margin. Trajectory generation uses a fifth-order polynomial interpolation method, ensuring zero velocity and acceleration at the start and end points, and setting the maximum joint velocity at intermediate points to 30° / s and the maximum acceleration to 60° / s. 2 It generates 100 trajectory points at 20ms intervals, with a total execution time of 2 seconds.

[0065] It should be noted that collision detection verification is a critical step to ensure safety. The system detected that at points 35 to 42 of the trajectory, the distance between the robotic arm's second link and the edge of the worktable was less than the safety threshold of 30mm, marking these as unsafe trajectory points. Using an artificial potential field method, a repulsive field was constructed around the obstacle, with its strength inversely proportional to the distance, guiding the robotic arm to avoid the obstacle and replanning a safe trajectory.

[0066] The Jacobian matrix J describes the effect of joint angle changes on the position of the end effector. It consists of the partial derivatives of the forward kinematics function:

[0067]

[0068] The Jacobian matrix is ​​a 3×n matrix, where n is the number of joints. Specifically:

[0069]

[0070] In this matrix, each column corresponds to a joint, and each row corresponds to a coordinate component (x, y, z) or an attitude component (e.g., rotation about the x, y, z axes) in Cartesian space.

[0071] Furthermore, mapping the current combination of joint angles to the end effector position in Cartesian space using a forward kinematics model includes:

[0072] p = f(q)

[0073] Where q is the joint angle vector, and p = [x, y, z] T Let T be the position coordinates of the end effector, T be the transpose, and f be the forward kinematic function, which is usually a nonlinear function that describes the relationship between the joint angle and the end effector position.

[0074] Furthermore, adjusting the angle value beyond the safe operating space boundary includes:

[0075]

[0076] Where, P n It is the angle adjustment value of the joint position of the robotic arm obtained after the nth correction, r represents the maximum radius from the geometric center of the obstacle to its surface, γ is the correction coefficient, A is the minimum distance between the joint position of the robotic arm and the geometric center of the obstacle, and B is the direction vector of the Euclidean distance.

[0077] This embodiment also provides an environment-aware robotic arm obstacle avoidance trajectory planning system, including: a joint space construction module, used to acquire joint angle data and end effector position data of the robotic arm, and to establish the joint space for the robotic arm's operation based on the joint angle data and end effector position data; an operation space extraction module, used to acquire environmental point cloud data, extract obstacle information based on the environmental point cloud data, and obtain the safe operation space of the robotic arm through the obstacle information; and an obstacle avoidance trajectory planning module, used to acquire the current joint angle combination of the robotic arm, convert the positional difference between the current joint angle combination and the target position into angle increments in the joint space, and adjust the angle values ​​exceeding the boundary of the safe operation space to complete the obstacle avoidance trajectory planning of the robotic arm.

[0078] Furthermore, the joint space construction module includes: a data acquisition unit for acquiring joint angle data of the robotic arm and position data of the end effector; and a joint space construction unit for converting the end effector position data into pose parameters in the Cartesian coordinate system and, in combination with the joint angle data, establishing the joint space in which the robotic arm operates.

[0079] Furthermore, the operation space extraction module includes: an obstacle information extraction unit, used to collect environmental point cloud data using a multi-target sensor composed of a depth camera and a lidar, and extract obstacle information based on the environmental point cloud data; and an operation space extraction unit, used to establish a transformation matrix between the multi-target sensor and the robotic arm based on the position of the multi-target sensor and the relationship between the robotic arm base, use the transformation matrix to spatially register the obstacle information, generate a unified 3D scene model containing the robotic arm and environmental objects, and calculate the minimum distance between each joint of the robotic arm and environmental objects through the unified 3D scene model. When the minimum distance is less than the safety threshold, it is marked as a potential collision risk area; when the minimum distance is greater than the safety threshold, it is marked as a safe operation space.

[0080] Furthermore, the obstacle avoidance trajectory planning module includes: a first obstacle avoidance trajectory planning unit, which maps the current joint angle combination to the end effector position in Cartesian space through a forward kinematics model, obtains the Euclidean distance between the current position of the end effector and the target position, constructs a Jacobian matrix using the partial derivatives of the forward kinematics model, handles the singular points of the Jacobian matrix, obtains a stable inverse Jacobian matrix, and uses the inverse stable inverse Jacobian matrix to convert the Euclidean distance into angle increments in joint space; and a second obstacle avoidance trajectory planning unit, which adjusts the angle values ​​that exceed the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm.

[0081] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for planning obstacle avoidance trajectory for a robotic arm based on environmental perception, characterized in that, include: Acquire the joint angle data and end effector position data of the robotic arm, and establish the joint space for the robotic arm's operation based on the joint angle data and end effector position data; Acquire environmental point cloud data, extract obstacle information based on the environmental point cloud data, and obtain the safe operating space of the robotic arm through the obstacle information; The current joint angle combination of the robotic arm is obtained, the positional difference between the current joint angle combination and the target position is converted into an angle increment in the joint space, and the angle value exceeding the boundary of the safe operating space is adjusted to complete the obstacle avoidance trajectory planning of the robotic arm. Adjusting the angle value beyond the boundary of the safe operating space includes: in, This is the first The angle adjustment value at the joint position of the robotic arm obtained from the second correction. This represents the maximum radius from the geometric center of the obstacle to its surface. For correction factor, This represents the minimum distance between the joint position of the robotic arm and the geometric center of the obstacle. is the direction vector of the Euclidean distance.

2. The obstacle avoidance trajectory planning method for a robotic arm based on environmental perception according to claim 1, characterized in that, Establishing the joint space includes: converting the end effector position data into pose parameters in a Cartesian coordinate system, and combining it with joint angle data to establish the joint space in which the robotic arm operates.

3. The obstacle avoidance trajectory planning method for a robotic arm based on environmental perception according to claim 1, characterized in that, Obtaining the safe operating space of the robotic arm includes: collecting environmental point cloud data using a multi-target sensor consisting of a depth camera and a lidar; extracting obstacle information based on the environmental point cloud data; establishing a transformation matrix between the multi-target sensor and the robotic arm based on the relationship between the position of the multi-target sensor and the robotic arm base; using the transformation matrix to spatially register the obstacle information to generate a unified 3D scene model containing the robotic arm and environmental objects; calculating the minimum distance between each joint of the robotic arm and environmental objects using the unified 3D scene model; marking a potential collision risk area when the minimum distance is less than a safety threshold, and marking a safe operating space when the minimum distance is greater than the safety threshold.

4. The obstacle avoidance trajectory planning method for a robotic arm based on environmental perception according to claim 1, characterized in that, Converting the positional difference between the current joint angle combination and the target position of the robotic arm into an angle increment in the joint space includes: mapping the current joint angle combination to the end effector position in Cartesian space using a forward kinematics model, obtaining the Euclidean distance between the current position of the end effector and the target position, constructing a Jacobian matrix using the partial derivatives of the forward kinematics model, processing the singular points of the Jacobian matrix to obtain a stable inverse Jacobian matrix, and using the inversely stable inverse Jacobian matrix to convert the Euclidean distance into an angle increment in the joint space.

5. The obstacle avoidance trajectory planning method for a robotic arm based on environmental perception according to claim 4, characterized in that, Mapping the current combination of joint angles to the end effector position in Cartesian space using the forward kinematics model includes: in, The joint angle vector. The position coordinates of the end effector. For transpose, Forward kinematics is a nonlinear function that describes the relationship between joint angles and end effector position.

6. A robotic arm obstacle avoidance trajectory planning system based on environmental perception, implemented according to any one of claims 1-5, characterized in that, include: The joint space construction module is used to acquire the joint angle data and end effector position data of the robotic arm, and to establish the joint space for the robotic arm to work based on the joint angle data and end effector position data. The operating space extraction module is used to acquire environmental point cloud data, extract obstacle information based on the environmental point cloud data, and obtain the safe operating space of the robotic arm through the obstacle information. The obstacle avoidance trajectory planning module is used to obtain the current joint angle combination of the robotic arm, convert the positional difference between the current joint angle combination and the target position into the angle increment in the joint space, and adjust the angle value that exceeds the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm.

7. The obstacle avoidance trajectory planning system for a robotic arm based on environmental perception according to claim 6, characterized in that, The joint space construction module includes: a data acquisition unit for acquiring joint angle data and end effector position data of the robotic arm; and a joint space construction unit for converting the end effector position data into pose parameters in a Cartesian coordinate system and, in conjunction with the joint angle data, establishing the joint space in which the robotic arm operates.

8. The obstacle avoidance trajectory planning system for a robotic arm based on environmental perception according to claim 6, characterized in that, The operation space extraction module includes: an obstacle information extraction unit, used to collect environmental point cloud data using a multi-target sensor composed of a depth camera and a lidar, and extract obstacle information based on the environmental point cloud data; and an operation space extraction unit, used to establish a transformation matrix between the multi-target sensor and the robotic arm base according to the position of the multi-target sensor and the relationship between the multi-target sensor and the robotic arm base, use the transformation matrix to spatially register the obstacle information, generate a unified three-dimensional scene model containing the robotic arm and environmental objects, and calculate the minimum distance between each joint of the robotic arm and the environmental objects through the unified three-dimensional scene model. When the minimum distance is less than a safety threshold, it is marked as a potential collision risk area; when the minimum distance is greater than the safety threshold, it is marked as a safe operation space.

9. The obstacle avoidance trajectory planning system for a robotic arm based on environmental perception according to claim 6, characterized in that, The obstacle avoidance trajectory planning module includes: a first obstacle avoidance trajectory planning unit, used to map the current joint angle combination to the end effector position in Cartesian space through a forward kinematics model, obtain the Euclidean distance between the current position of the end effector and the target position, construct a Jacobian matrix using the partial derivatives of the forward kinematics model, process the singular points of the Jacobian matrix to obtain a stable inverse Jacobian matrix, and use the inverse stable inverse Jacobian matrix to convert the Euclidean distance into an angle increment in joint space; and a second obstacle avoidance trajectory planning unit, used to adjust the angle values ​​that exceed the boundary of the safe operating space to complete the obstacle avoidance trajectory planning of the robotic arm.

Images