A framework structure for hierarchical humanoid control of unmanned equipment platforms

Abstract

The present invention belongs to the field of robot motion control technology, and specifically relates to a framework structure for hierarchical humanoid control of unmanned equipment platforms, which is divided into a perception unit and a control unit. The humanoid control framework establishes local elevation, semantics and environmental models by acquiring positioning and depth perception information, and at the same time establishes two services of coarse and fine granularity under the drive of task instructions, completes the selection and adjustment of various control strategies and unit parameters in the control unit, and achieves the best comprehensive computing power and performance; the control unit uses the perception unit planning as the expected trajectory, realizes tracking control based on body stability and limb planning, and finally completes the solution of servo instructions. Compared with the prior art, the present invention adopts hierarchical drive and humanoid control bionics to design the control architecture of unmanned systems and platforms, solving the problems of complex design, coupling, and difficult system update and maintenance of traditional control systems.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot motion control, and in particular relates to a framework structure for hierarchical humanoid control of an unmanned equipment platform. Background Art

[0002] Traditional unmanned equipment mainly adopts the idea of ​​centralized control to design control systems. There are also distributed control solutions, but in general, fixed input and output control methods are used to control unmanned equipment. Unmanned equipment is often used in complex environments and multi-dynamic tasks, especially in fully autonomous unmanned operations that will be used in future operations. Unmanned equipment needs to comprehensively improve its autonomous decision-making and environmental perception capabilities, and further achieve more dynamic, high-precision stable control. In order to achieve the above goals and further combine traditional control strategies with advanced intelligent algorithms, a new control framework needs to be built. Summary of the invention

[0003] 1. Technical issues to be resolved

[0004] The technical problem to be solved by the present invention is: how to propose a framework structure for hierarchical humanoid control of unmanned equipment platforms, which adopts the principle of humanoid intelligent control to establish a hierarchical, coarse-grained and iteratively optimized humanoid control framework, thereby improving the motion control capabilities of unmanned equipment in complex environments and variable tasks.

[0005] (II) Technical solution

[0006] In order to solve the above technical problems, the present invention provides a framework structure for hierarchical human-like control of unmanned equipment platforms, wherein the framework structure for hierarchical human-like control of unmanned equipment platforms comprises: a perception unit and a control unit;

[0007] The perception unit includes: a global positioning module, a perception mapping module, an elevation semantic fusion module, a target recognition module, a target behavior understanding module, an ontology instruction planning module, and a perception unit;

[0008] The global positioning module is used to realize data fusion based on the GPS global positioning system and SINS integrated navigation system carried by the unmanned equipment through the integrated navigation algorithm, solve the global positioning information of the unmanned equipment platform, further perform online fault diagnosis on the global positioning information, and judge the health of the GPS global positioning system and the SINS integrated navigation system;

[0009] The perception mapping module is used to match the LiDAR radar and RGBD camera point clouds, and generate a global occupancy map and a local elevation map in combination with global positioning information;

[0010] The elevation semantic fusion module is used to extract environmental macro-parameter information from the global occupancy map and the local elevation map using semantic analysis and digital terrain GIS analysis technology, and generate a GIS terrain semantic local elevation map;

[0011] The target recognition module is used to extract features from LiDAR radar and RGBD camera images and generate corresponding label classification and target pose estimation results;

[0012] The target behavior understanding module is used to analyze the target behavior and predict the target motion prediction trajectory in the future based on the label classification and target pose estimation results;

[0013] The ontology instruction planning module is used to obtain global positioning information, global occupancy map, GIS terrain semantic local elevation map, environmental macro parameter information and target predicted motion trajectory, and complete the planning of its coarse-grained ontology motion trajectory and control mode in combination with the simplified dynamic model of the unmanned equipment; the ontology instruction planning module finally inputs the above-generated coarse-grained ontology motion trajectory and control mode planning, GIS terrain semantic local elevation map and target motion predicted trajectory into the shared memory;

[0014] The control unit includes: a body state estimation module, a limb motion planning module, a body stability control module, a terminal servo control module, a coarse-grained evolution service module, and a fine-grained evolution service module; the control unit obtains perception data from a shared memory to complete intelligent control;

[0015] The body state estimation module is used to obtain terrain data from the shared memory, combine IMU and encoder data to complete the judgment of the support state, and finally select the Kalman filter KF or the extended Kalman filter EKF to fuse the data based on the coarse-grained evolution service module to estimate the motion state of the body;

[0016] The limb motion planning module is used to obtain terrain data from the shared memory, complete the selection of the landing point and the planning of the limb swing trajectory in combination with the main body motion state estimation, and finally complete the adjustment of the swing parameter adaptation and the landing point optimization reward parameter based on the fine-grained evolution service module;

[0017] The body stability control module is used to obtain the planning of the coarse-grained body motion trajectory and control mode from the shared memory, and selects VMC virtual servo control or MPC model predictive controller under the drive of the coarse-grained evolution service module to complete the calculation of the reaction force of the support point in combination with the motion state estimation of the body; further, the distribution of the support force is completed based on the QP quadratic optimization allocation algorithm, and the adjustment of different control weights is realized under the drive of the fine-grained evolution service module;

[0018] The peripheral servo control module is used to calculate the actuator position, speed and torque instructions to complete the limb swing trajectory tracking and the closed-loop tracking of the support point reaction force, and finally output the corresponding instructions through the servo bus. At the same time, under the drive of the fine-grained evolution service module, it completes the adjustment of the position, speed and torque tracking controller gain and damping coefficient and channel switching.

[0019] Among them, the coarse-grained evolution service module is used to realize the selection of corresponding algorithm units in the entity state estimation and entity stability control modules under the joint drive of tasks and the environmental model constructed by GIS and elevation maps, so as to achieve the optimal computing power requirement and performance of the overall system.

[0020] Among them, the fine-grained evolution service module is used to complete the selection of position, speed and torque control channels based on the adaptive control strategy under the drive of servo command tracking error, and at the same time complete the adjustment of the stiffness and damping coefficient of each channel.

[0021] Wherein, the perception data includes coarse-grained ontological motion trajectory.

[0022] Wherein, the perception data includes planning of the control mode.

[0023] Wherein, the perception data includes GIS terrain semantic local elevation map.

[0024] Wherein, the perception data includes a predicted target motion trajectory.

[0025] Wherein, the perception data includes terrain data.

[0026] Among them, in order to solve the problems of complex and difficult design of the overall control system of traditional unmanned platforms and complex system coupling, the framework structure of hierarchical human-like control for unmanned equipment platforms is proposed.

[0027] Among them, the framework structure of hierarchical humanoid control for unmanned equipment platforms is divided into a perception unit and a control unit. The humanoid control framework establishes local elevation, semantic and environmental models by acquiring positioning and depth perception information, and at the same time establishes coarse and fine granularity services driven by task instructions to complete the selection and adjustment of various control strategies and unit parameters in the control unit, so as to achieve the best comprehensive computing power and performance; the control unit uses the perception unit planning as the expected trajectory, realizes tracking control based on body stability and limb planning, and finally completes the solution of servo instructions.

[0028] (III) Beneficial effects

[0029] The present invention proposes a framework structure for hierarchical humanoid control of unmanned equipment platforms, which adopts the principle of humanoid intelligent control to establish a hierarchical, coarse-grained and iteratively optimized humanoid control framework, thereby improving the motion control capability of unmanned equipment in complex environments and with variable tasks.

[0030] Compared with the prior art, the hierarchical humanoid control framework of the unmanned equipment platform of the present invention divides the unmanned equipment control system from coarse and fine granularity, simplifies the control logic design and is conducive to interface standardization; by introducing coarse and fine granularity evolution services, the platform can generate autonomous strategy switching according to task and environmental changes, and realize dynamic adjustment of control unit parameters, thereby improving the intelligence of the control system. The control architecture of unmanned systems and platforms is designed by adopting hierarchical drive and humanoid control bionics, which solves the problems of complex design, coupling, and difficult system update and maintenance of traditional control systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the framework structure of the hierarchical human-like control of unmanned equipment platforms according to the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, content, and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below in conjunction with the accompanying drawings and examples.

[0033] In order to solve the above technical problems, the present invention provides a framework structure for hierarchical human-like control of unmanned equipment platforms, such as Figure 1 As shown, the framework structure for hierarchical humanoid control of unmanned equipment platforms includes: a perception unit and a control unit;

[0034] The perception unit includes: a global positioning module, a perception mapping module, an elevation semantic fusion module, a target recognition module, a target behavior understanding module, an ontology instruction planning module, and a perception unit;

[0035] The global positioning module is used to realize data fusion based on the GPS global positioning system and SINS integrated navigation system carried by the unmanned equipment through the integrated navigation algorithm, solve the global positioning information of the unmanned equipment platform, further perform online fault diagnosis on the global positioning information, and judge the health of the GPS global positioning system and the SINS integrated navigation system;

[0036] The perception mapping module is used to match the LiDAR radar and RGBD camera point clouds, and generate a global occupancy map and a local elevation map in combination with global positioning information;

[0037] The elevation semantic fusion module is used to extract environmental macro-parameter information from the global occupancy map and the local elevation map using semantic analysis and digital terrain GIS analysis technology, and generate a GIS terrain semantic local elevation map;

[0038] The target recognition module is used to extract features from LiDAR radar and RGBD camera images and generate corresponding label classification and target pose estimation results;

[0039] The target behavior understanding module is used to analyze the target behavior and predict the target motion prediction trajectory in the future based on the label classification and target pose estimation results;

[0040] The ontology instruction planning module is used to obtain global positioning information, global occupancy map, GIS terrain semantic local elevation map, environmental macro parameter information and target predicted motion trajectory, and complete the planning of its coarse-grained ontology motion trajectory and control mode in combination with the simplified dynamic model of the unmanned equipment; the ontology instruction planning module finally inputs the above-generated coarse-grained ontology motion trajectory and control mode planning, GIS terrain semantic local elevation map and target motion predicted trajectory into the shared memory;

[0041] The control unit includes: a body state estimation module, a limb motion planning module, a body stability control module, a terminal servo control module, a coarse-grained evolution service module, and a fine-grained evolution service module; the control unit obtains perception data from a shared memory to complete intelligent control;

[0042] The body state estimation module is used to obtain terrain data from the shared memory, combine IMU and encoder data to complete the judgment of the support state, and finally select the Kalman filter KF or the extended Kalman filter EKF to fuse the data based on the coarse-grained evolution service module to estimate the motion state of the body;

[0043] The limb motion planning module is used to obtain terrain data from the shared memory, complete the selection of the landing point and the planning of the limb swing trajectory in combination with the main body motion state estimation, and finally complete the adjustment of the swing parameter adaptation and the landing point optimization reward parameter based on the fine-grained evolution service module;

[0044] The body stability control module is used to obtain the planning of the coarse-grained body motion trajectory and control mode from the shared memory, and selects VMC virtual servo control or MPC model predictive controller under the drive of the coarse-grained evolution service module to complete the calculation of the reaction force of the support point in combination with the motion state estimation of the body; further, the distribution of the support force is completed based on the QP quadratic optimization allocation algorithm, and the adjustment of different control weights is realized under the drive of the fine-grained evolution service module;

[0045] The peripheral servo control module is used to calculate the actuator position, speed and torque instructions to complete the limb swing trajectory tracking and the closed-loop tracking of the support point reaction force, and finally output the corresponding instructions through the servo bus. At the same time, under the drive of the fine-grained evolution service module, it completes the adjustment of the position, speed and torque tracking controller gain and damping coefficient and channel switching.

[0046] Among them, the coarse-grained evolution service module is used to realize the selection of corresponding algorithm units in the entity state estimation and entity stability control modules under the joint drive of tasks and the environmental model constructed by GIS and elevation maps, so as to achieve the optimal computing power requirement and performance of the overall system.

[0047] Among them, the fine-grained evolution service module is used to complete the selection of position, speed and torque control channels based on the adaptive control strategy under the drive of servo command tracking error, and at the same time complete the adjustment of the stiffness and damping coefficient of each channel.

[0048] Wherein, the perception data includes coarse-grained ontological motion trajectory.

[0049] Wherein, the perception data includes planning of the control mode.

[0050] Wherein, the perception data includes GIS terrain semantic local elevation map.

[0051] Wherein, the perception data includes a predicted target motion trajectory.

[0052] Wherein, the perception data includes terrain data.

[0053] Among them, in order to solve the problems of complex and difficult design of the overall control system of traditional unmanned platforms and complex system coupling, the framework structure of hierarchical human-like control for unmanned equipment platforms is proposed.

[0054] Among them, the framework structure of hierarchical humanoid control for unmanned equipment platforms is divided into a perception unit and a control unit. The humanoid control framework establishes local elevation, semantic and environmental models by acquiring positioning and depth perception information, and at the same time establishes coarse and fine granularity services driven by task instructions to complete the selection and adjustment of various control strategies and unit parameters in the control unit, so as to achieve the best comprehensive computing power and performance; the control unit uses the perception unit planning as the expected trajectory, realizes tracking control based on body stability and limb planning, and finally completes the solution of servo instructions.

[0055] Example 1

[0056] In this embodiment, the provided unmanned equipment platform hierarchical humanoid control framework is divided into a perception unit and a control unit. The humanoid control framework establishes local elevation, semantic and environmental models by acquiring positioning and depth perception information, and at the same time establishes two services of coarse and fine granularity driven by task instructions, completes the selection and adjustment of each control strategy and unit parameter in the control unit, and achieves the best comprehensive computing power and performance; the control unit uses the perception unit planning as the expected trajectory, realizes tracking control based on body stability and limb planning, and finally completes the solution of the servo instruction. The specific design process of each unit is as follows:

[0057] Perception Unit:

[0058] (1) The perception unit is divided into three modules: global positioning, perception mapping, and target recognition. The global positioning module is based on the GPS global positioning system and the SINS integrated navigation system carried by the unmanned equipment. It realizes data fusion through the integrated navigation algorithm and solves the global positioning information of the unmanned equipment platform. Furthermore, through the fault diagnosis algorithm and the internal feedback status of the sensor, it realizes online diagnosis of the global positioning system fault, judges the health of the GPS global positioning system and the SINS integrated navigation system, and makes corresponding fusion channel switching and weight adjustment.

[0059] (2) The perception and mapping module matches the LiDAR radar and RGBD camera point clouds, uses the radar to build a large-scale global occupancy map, and uses the RGBD camera dense point cloud combined with the global positioning results to build a local elevation map.

[0060] (3) The elevation semantic fusion module uses semantic analysis and digital terrain GIS analysis technology to extract environmental macro-parameter information from the global occupancy map and the local elevation map, and analyzes the ruggedness, slope, large drop edge and undulation of the current terrain; the above information is superimposed on the local elevation map in a semantic way to establish a GIS terrain semantic local elevation map;

[0061] (4) Target recognition module: It extracts features from LiDAR radar and RGBD camera images, identifies characteristic targets such as people and vehicles, generates corresponding label classifications and local positioning results relative to the perception system, and further combines global positioning information to complete the position and posture conversion of the target in the global coordinate system;

[0062] (5) Target behavior understanding module, based on label classification and target pose estimation results, completes the analysis of its behavior and predicts its future motion trajectory;

[0063] (6) The main body instruction planning module obtains global positioning information, global occupancy map, local elevation map, environmental macro parameter information and target predicted motion trajectory, completes the planning of the main body track waypoints, further combines the simplified dynamic model of the unmanned equipment to complete the online planning of its continuous coarse-grained motion trajectory, and determines the control mode of the platform in each trajectory segment;

[0064] (7) Finally, the coarse-grained ontology motion trajectory, control mode, GIS terrain semantic local elevation map, global positioning information, and target motion prediction trajectory generated by the above modules in real time are input into the shared memory to complete the real-time interaction with the control unit data;

[0065] Control unit:

[0066] (1) The control unit obtains the perception data from the shared memory, which mainly includes body state estimation, limb motion planning, body stability control, peripheral servo control, and coarse-grained evolution and fine-grained evolution services. It completes the tracking of the trajectory instructions planned by the perception unit and completes the solution of the servo drive instructions;

[0067] (2) The body state estimation module obtains terrain data from the shared memory, combines IMU and encoder data, completes the judgment of the support state of the drive execution unit, and finally selects the Kalman filter KF (low precision, low computing power) or extended Kalman filter EKF (high precision, high computing power) to fuse the data based on the coarse-grained evolution service to estimate the body's motion state such as position, velocity, acceleration and support force;

[0068] (3) Limb motion planning module, which obtains terrain data from shared memory, completes the selection of support points in combination with the estimation results of the body's motion state, further realizes the planning of the limb swing trajectory, and finally completes the adaptation of the swing trajectory planning parameters and the adjustment of the weights in the support point selection reward function based on the fine-grained evolution service;

[0069] (4) The body stability control module obtains the coarse-grained motion trajectory planning and control mode from the shared memory, and selects VMC virtual servo control (low computing power, low performance) or MPC model predictive controller (high computing power, high performance) under the coarse-grained evolution service drive in combination with the body state estimation to complete the calculation of the support point reaction force; further, the support force is allocated based on the QP quadratic optimization allocation algorithm, and different allocation weights and control gain parameters are adjusted under the fine-grained service drive;

[0070] (5) The terminal servo control module completes the calculation of the position, speed and torque of the servo actuator, and drives the platform to track the limb swing trajectory and the reaction force of the support point. The drive command is output to the actuator through the servo bus; at the same time, driven by the fine-grained evolution service, based on the closed-loop error of position, speed and force, the position, speed and torque tracking controller gain and damping coefficient are adjusted online and the channel is switched;

[0071] Coarse-grained and fine-grained evolution services:

[0072] (1) Coarse-grained evolution service: The coarse-grained evolution service is aimed at the body state estimation and body stability control modules, and mainly realizes the selection of control strategies and models. Based on the current task and control mode, it combines the local elevation map and the environment model constructed by GIS semantic information, and switches the algorithm unit under the joint drive of the two to achieve the comprehensive optimization of the control system in terms of computing power and performance.

[0073] (2) Fine-grained evolutionary service: Driven by the limb position, velocity, and support force tracking errors, the switching selection of the position, velocity, and torque control channels is completed based on the adaptive control strategy, and the stiffness and damping coefficients of each channel are adaptively adjusted to improve the control compliance and accuracy.

[0074] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A framework structure for hierarchical humanoid control of unmanned equipment platforms, characterized in that: The framework structure for hierarchical humanoid control of unmanned equipment platforms includes: a perception unit and a control unit; The perception unit includes: a global positioning module, a perception mapping module, an elevation semantic fusion module, a target recognition module, a target behavior understanding module, and an ontology instruction planning module; The global positioning module is used to realize data fusion based on the GPS global positioning system and SINS integrated navigation system carried by the unmanned equipment through the integrated navigation algorithm, solve the global positioning information of the unmanned equipment platform, perform online fault diagnosis on the global positioning information, and judge the health of the GPS global positioning system and the SINS integrated navigation system; The perception mapping module is used to match the LiDAR radar and RGBD camera point clouds, and generate a global occupancy map and a local elevation map in combination with global positioning information; The elevation semantic fusion module is used to extract environmental macro-parameter information from the global occupancy map and the local elevation map using semantic analysis and digital terrain GIS analysis technology, and generate a GIS terrain semantic local elevation map; The target recognition module is used to extract features from LiDAR radar and RGBD camera images and generate corresponding label classification and target pose estimation results; The target behavior understanding module is used to analyze the target behavior and predict the target motion prediction trajectory in the future based on the label classification and target pose estimation results; The ontology instruction planning module is used to obtain global positioning information, global occupancy map, GIS terrain semantic local elevation map, environmental macro parameter information and target predicted motion trajectory, and complete the planning of its coarse-grained ontology motion trajectory and control mode in combination with the simplified dynamic model of the unmanned equipment; the ontology instruction planning module finally inputs the above-generated coarse-grained ontology motion trajectory and control mode planning, GIS terrain semantic local elevation map and target motion predicted trajectory into the shared memory; The control unit includes: a body state estimation module, a limb motion planning module, a body stability control module, a terminal servo control module, a coarse-grained evolution service module, and a fine-grained evolution service module; the control unit obtains perception data from a shared memory to complete intelligent control; The body state estimation module is used to obtain terrain data from the shared memory, combine IMU and encoder data to complete the judgment of the support state, and finally select the Kalman filter KF or the extended Kalman filter EKF to fuse the data based on the coarse-grained evolution service module to estimate the motion state of the body; The limb motion planning module is used to obtain terrain data from the shared memory, complete the selection of the landing point and the planning of the limb swing trajectory in combination with the main body motion state estimation, and finally complete the adjustment of the swing parameter adaptation and the landing point optimization reward parameter based on the fine-grained evolution service module; The body stability control module is used to obtain the coarse-grained body motion trajectory and control mode planning from the shared memory, and selects VMC virtual servo control or MPC model predictive controller under the drive of the coarse-grained evolution service module to complete the calculation of the support point reaction force; completes the distribution of the support force based on the QP quadratic optimization allocation algorithm, and realizes the adjustment of different control weights under the drive of the fine-grained evolution service module; The peripheral servo control module is used to calculate the actuator position, speed and torque instructions to complete the limb swing trajectory tracking and the closed-loop tracking of the support point reaction force, and finally output the corresponding instructions through the servo bus. At the same time, under the drive of the fine-grained evolution service module, it completes the adjustment of the position, speed and torque tracking controller gain and damping coefficient and channel switching.

2. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The coarse-grained evolution service module is used to realize the selection of corresponding algorithm units in the entity state estimation and entity stability control modules under the joint drive of tasks and the environmental model constructed by GIS and elevation maps, so as to achieve the optimal computing power requirement and performance of the overall system.

3. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 2, characterized in that: The fine-grained evolution service module is used to complete the selection of position, speed and torque control channels based on the adaptive control strategy under the drive of the servo instruction tracking error, and to complete the adjustment of the stiffness and damping coefficient of each channel.

4. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The perception data includes coarse-grained body motion trajectories.

5. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The sensory data includes a plan for a control mode.

6. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The perception data includes a GIS terrain semantic local elevation map.

7. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The perception data includes a predicted target motion trajectory.

8. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The sensory data includes terrain data.

9. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: In order to solve the problems of complex and difficult design of the overall control system of traditional unmanned platforms and complex system coupling, a framework structure for hierarchical humanoid control of unmanned equipment platforms is proposed.

10. The framework structure for hierarchical humanoid control of unmanned equipment platforms as claimed in claim 1, characterized in that: The framework structure of hierarchical humanoid control for unmanned equipment platforms is divided into a perception unit and a control unit. The humanoid control framework establishes local elevation, semantic and environmental models by acquiring positioning and depth perception information, and at the same time establishes coarse and fine granularity services driven by task instructions to complete the selection and adjustment of various control strategies and unit parameters in the control unit, thereby achieving the best comprehensive computing power and performance. The control unit uses the perception unit planning as the expected trajectory, realizes tracking control based on body stability and limb planning, and finally completes the solution of the servo instructions.

Images