CN-121764122-B - Artificial intelligence autonomous navigation and path planning control system based on edge calculation

CN121764122BCN 121764122 BCN121764122 BCN 121764122BCN-121764122-B

Abstract

The invention relates to the technical field of mobile robots and intelligent control, in particular to an artificial intelligent autonomous navigation and path planning control system based on edge calculation, which comprises an environment sensing step, a tensor field modulation step, a geodesic line flow control step and a driving step, wherein the environment sensing step is used for collecting environment point cloud data and voicing to generate local environment field data containing space distribution information, the tensor field modulation step is used for mapping the local environment field data into a measurement tensor field in a Riemann manifold space to construct a measurement structure for representing environmental resistance characteristics, the geodesic line flow control step is used for solving a geodesic line differential equation based on the measurement tensor field to directly generate a speed vector control instruction for representing acceleration, and the driving step is executed to convert the response instruction into a motor control signal to regulate and output.

Inventors

HE YIFAN
SU JINHE
WANG ZONGYUE

Assignees

集美大学

Dates

Publication Date: 20260505
Application Date: 20260305

Claims (7)

1. The artificial intelligent autonomous navigation and path planning control system based on edge calculation is characterized by comprising an environment sensing unit, a processing unit and a processing unit, wherein the environment sensing unit is configured at the edge side of a mobile robot and is used for acquiring environment point cloud data, acquiring global coordinates of a target point, and carrying out voxelization on the environment point cloud data to generate local environment field data containing space distribution information; The edge tensor field modulator is used for receiving the local environment field data, mapping the local environment field data into a measurement tensor field in the Riemann manifold space in real time, and constructing a measurement structure representing the environmental resistance characteristic by calculating the geometric attribute of the local space; The geodesic flow controller is used for directly generating a speed vector control instruction by solving a geodesic differential equation based on the measurement tensor field and the global coordinates of the target point, wherein the speed vector control instruction is an acceleration vector representing a path along the geodesic; the execution driving unit is used for responding to the speed vector control instruction, converting the speed vector control instruction into a motor control signal of a bottom layer and adjusting the motor output of the mobile robot; the environment sensing unit includes: The data voxelization module is used for dividing the acquired original environment point cloud data into three-dimensional voxel grids, and calculating the point cloud density in each voxel without carrying out binary occupation judgment on the voxels; And the potential energy encoding module is used for distributing rejection potential energy values to each voxel according to the point cloud density, wherein the rejection potential energy values and the point cloud density are in positive correlation, voxel field data with potential energy information are generated, and the voxel field data are transmitted to the edge tensor field modulator.
2. The edge computing-based artificial intelligence autonomous navigation and path planning control system of claim 1, wherein the edge tensor field modulator comprises: the measurement tensor generation module is used for converting the repulsive potential energy value in the voxel field data into measurement tensor disturbance of local space, generating a high-curvature space structure in a region with the obstacle density higher than a preset density threshold value, and maintaining a flat Euclidean measurement structure in a region with the obstacle density not higher than the preset density threshold value; And the symbol calculation module is used for calculating the Cristofer symbol of the current local space in real time by utilizing the hardware acceleration capability of the edge calculation node based on the generated measurement tensor field, wherein the Cristofer symbol is used for describing the connection coefficient in the Riemann manifold space.
3. The edge computing-based artificial intelligence autonomous navigation and path planning control system of claim 2, wherein the edge tensor field modulator is further configured with: The dynamic environment processing module is used for processing dynamic change data in the environment, identifying moving obstacles, and updating curvature parameters of corresponding areas in the measurement tensor field in real time according to the speed and the direction of the dynamic obstacles so that the measurement structure dynamically changes along with the environment.
4. The edge computation based artificial intelligence autonomous navigation and path planning control system of claim 1, wherein the geodetic wire flow controller comprises: The equation solving module is used for constructing and solving a geodesic differential equation, taking the geometric curvature information output by the edge tensor field modulator as constraint conditions of the equation, and calculating a local gradient direction pointing to the target point position from the current position in the current Riemann manifold space; And the vector synthesis module is used for synthesizing a virtual gravity vector for driving the robot according to the local gradient direction, wherein the virtual gravity vector directly corresponds to the motion acceleration of the robot, and a discrete path coordinate point set is not required to be generated.
5. The edge computation based artificial intelligence autonomous navigation and path planning control system of claim 4, wherein the geodetic wire flow controller further comprises: The implicit speed self-adaptive module is used for automatically adjusting the output speed according to the space curvature change of the measurement tensor field and calculating the measurement distance on the current path, outputting a speed vector control instruction for reducing the amplitude value if the space curvature is higher than a preset curvature threshold value, and outputting a speed vector control instruction for maintaining or increasing the amplitude value if the space curvature is not higher than the curvature threshold value.
6. The edge computing-based artificial intelligence autonomous navigation and path planning control system of claim 1, wherein the execution driving unit comprises: The signal mapping module is used for receiving the speed vector control instruction and decomposing the acceleration vector into a linear speed component and an angular speed component; And the pulse width modulation generation module is used for calculating the voltage duty ratio required by the servo motor according to the linear velocity component and the angular velocity component, generating PWM voltage signals and executing steering and acceleration and deceleration actions by driving the hub motor.
7. The edge computation-based artificial intelligence autonomous navigation and path planning control system according to claim 4, wherein the edge tensor field modulator and the geodetic wire flow controller are configured in an FPGA hardware accelerator or GPU parallel computing unit, and the matrix operation of the metric tensor field and the gradient solution of the geodetic wire differential equation are processed in real time through a parallel pipeline architecture, so that the processing delay from the acquisition of the environmental point cloud data to the generation of the speed vector control instruction is controlled within a preset time threshold.

Description

Artificial intelligence autonomous navigation and path planning control system based on edge calculation Technical Field The invention relates to the technical field of mobile robots and intelligent control, in particular to an artificial intelligent autonomous navigation and path planning control system based on edge calculation. Background In the technical field of position or track control of the current non-rail guided vehicle, when a mobile robot works in an unstructured environment, the mobile robot needs to rely on an edge side sensor to acquire environment point cloud data so as to perform autonomous navigation; The existing G05D navigation scheme generally adopts a serial computing architecture of a sensing, planning and control layer, namely, environmental data is firstly processed into a binary occupied grid map, a discrete global path point set is generated based on the binary occupied grid map, and then a motor is driven to execute the discrete global path point set through a track tracking algorithm. Although the scheme has basic traffic capacity in a static low-speed scene, the system has a remarkable serial calculation bottleneck due to the fact that the sensing and the control links are separated and the flow is long. In addition, the binary map ignores the spatial density distribution characteristics of the obstacles, and the discrete path planning is difficult to adapt to a dynamically-changing complex environment, so that the response delay is high in the navigation process, the dynamic obstacle avoidance capability is weak, and the real-time safety control requirement under a dense people stream scene is difficult to meet. Therefore, how to break through the serial limitation of the traditional perception planning control, realize the rapid mapping of the environment perception action execution under the limited condition of the edge computing resource, and promote the real-time response speed of the system to the dynamic obstacle and the smoothness of the path planning are technical problems to be solved. Disclosure of Invention In order to solve the technical problems, the invention provides an artificial intelligent autonomous navigation and path planning control system based on edge calculation, and specifically, the technical scheme of the invention comprises the following steps: The environment sensing unit is configured at the edge side of the mobile robot and is used for acquiring environment point cloud data, acquiring global coordinates of a target point, voxelizing the environment point cloud data and generating local environment field data containing space distribution information; the edge tensor field modulator is used for receiving local environment field data, mapping the local environment field data into a measurement tensor field in the Riemann manifold space in real time, and constructing a measurement structure representing the environmental resistance characteristic by calculating the geometric property of the local space; The geodesic flow controller is used for directly generating a speed vector control instruction by solving a geodesic differential equation based on the measurement tensor field and the global coordinates of the target point, wherein the speed vector control instruction is used for representing an acceleration vector along a geodesic path; And the execution driving unit is used for responding to the speed vector control command, converting the speed vector control command into a motor control signal of the bottom layer and adjusting the motor output of the mobile robot. Preferably, the environment sensing unit includes: The data voxelization module is used for dividing the acquired original environment point cloud data into three-dimensional voxel grids, and calculating the point cloud density in each voxel without carrying out binary occupation judgment on the voxels; the potential energy encoding module is used for distributing rejection potential energy values to each voxel according to the point cloud density, wherein the rejection potential energy values and the point cloud density are in positive correlation, voxel field data with potential energy information are generated, and the voxel field data are transmitted to the edge tensor field modulator. Preferably, the edge tensor field modulator includes: The measurement tensor generation module is used for converting repulsive potential energy values in the voxel field data into measurement tensor disturbance of local space, generating a high-curvature space structure in a region with the obstacle density higher than a preset density threshold value, and maintaining a flat Euclidean measurement structure in a region with the obstacle density not higher than the preset density threshold value; And the symbol calculation module is used for calculating the Kristofer symbol of the current local space in real time by utilizing the hardware acceleration capability of the edge calculation node based