Aerial robot perception uncertain motion planning method, system and storage medium

Abstract

The application provides a kind of flying robot perception uncertain motion planning method, system and storage medium, belong to robot motion planning field.To solve the problem that the trajectory planning of traditional robot is difficult to meet the geometric constraint and the task constraint about screen pointing and low noise, and difficult to avoid obstacle in real time in small space.The application includes that robot uses improved A* algorithm for global path planning according to three-dimensional grid map;Robot uses Tube-MPC method for local path planning according to the initial path generated by global path planning and the motion information of astronauts and floating objects in cabin and obtains control quantity.The application uses improved A* to improve path search efficiency, facilitates astronaut to check screen information, and makes robot adjust attitude as far as possible using flywheel, reduces noise brought by ducted fan;Tube-MPC is used for back-end trajectory optimization, and robust motion trajectory meeting relevant constraint requirements is obtained.

Description

Technical Field

[0001] This invention relates to the field of robot motion planning technology, and more specifically, to a method, system, and storage medium for motion planning of a flying robot that perceives uncertainty. Background Technology

[0002] In the process of space exploration and space experiment development, robots with autonomous perception, decision-making, planning, and control capabilities play a crucial role. During the construction and operation of my country's space station, astronauts need to conduct various space experiments and maintain equipment. There is an urgent need for intelligent space robots operating in microgravity environments to assist astronauts in tasks such as tracking and lighting, filming, equipment monitoring and repair, tool transfer, equipment disassembly and replacement, and scientific experiment care, thereby improving astronauts' work efficiency.

[0003] Autonomous motion planning technology for robots in uncertain environments is crucial to ensuring the platform achieves the aforementioned functions. In the fields of drones and unmanned vehicles, many mature path planning algorithms already exist. However, for the specific nonholonomic constraints of the dynamic model of intelligent space robots in microgravity environments, and their unique unstructured application scenarios and task requirements, algorithmic improvements are needed in both global and local path planning.

[0004] In global path planning, the traditional A* algorithm aims to quickly plan a path from the starting point to the destination while satisfying the geometric constraints of space. However, it doesn't consider the robot's kinematics and dynamics, which may result in an impassable path. Furthermore, the continuous maintenance and access to unnecessary neighboring nodes during the search process increases computational load and reduces algorithm efficiency. Unlike other robots, intelligent space robots in microgravity environments need to provide stable human-machine interaction and generate the lowest possible maneuvering noise (in microgravity environments, intelligent space robots collaborate with astronauts via an onboard touchscreen, requiring consideration of screen pointing constraints during maneuvers. Given the long-term coexistence of the robot and astronauts in a confined space, minimizing maneuvering noise is crucial). Therefore, its global motion planning algorithm should not only satisfy geometric constraints but also task constraints such as screen pointing and low noise.

[0005] In terms of local path planning, in addition to optimization operations such as trajectory smoothing, space intelligent flight robots in microgravity environments also need to replan their corresponding trajectories in real time based on predicted information such as the movement direction and speed of astronauts and floating objects inside the cabin, in order to avoid dynamic obstacles or track astronauts in real time. Furthermore, existing backend optimization methods do not consider uncertainties in perception, mapping, and control, and only perform trajectory optimization based on the robot's ideal state-space model, which may result in poor performance in practical applications. Summary of the Invention

[0006] The technical problem to be solved by this invention is:

[0007] To address the problems that traditional robot trajectory planning is not suitable for microgravity environments, it is difficult to meet geometric constraints and task constraints regarding screen orientation and low noise, and it is difficult to avoid obstacles in small spaces in real time.

[0008] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0009] This invention provides a method for flight robot to perceive uncertain motion planning, comprising the following steps:

[0010] S100. The robot performs global path planning using an improved A* algorithm based on a 3D grid map. The improved A* algorithm is based on the traditional A* algorithm, but with node expansion and evaluation function optimization, to obtain the trajectory points generated by the global path planning.

[0011] S200: Based on the trajectory points generated by the global path planning in step S100, the robot performs local path planning using the motion information of the astronauts and floating objects inside the cabin, kinematic and dynamic constraints, and Tube constraints. Through model predictive control, the actual safety conditions are modeled as inequality constraints of an optimization problem, the uncertainties of perception and model are transformed into Tube constraints, and the goals of accuracy and energy consumption are modeled as objective functions for evaluating the motion trajectory. By solving for the optimal value of the objective function in the prediction time domain and its corresponding motion trajectory, the robot's motion trajectory planning is achieved.

[0012] Further, in step S100, the following is included:

[0013] The S110 airborne core processor first generates positioning data based on the binocular images collected by the forward-looking obstacle avoidance camera. Then, it fuses the positioning data with the position data collected by the rear-looking navigation camera and the laser ranging data to obtain higher-precision robot position data. Based on the robot position data, combined with the forward-looking depth point cloud image and the binocular fisheye stereo vision image of the rear-looking navigation camera, the TSDF method is used to map the surrounding environment.

[0014] S120. The robot uses an improved A* algorithm to perform global path planning based on a 3D grid map, including node expansion and evaluation function optimization.

[0015] Further, in step S120, the following is included:

[0016] The evaluation function f(n) of the S121 and A* algorithms is expressed as:

[0017]

[0018] In the formula, f(n) represents the cost function from the starting point to the target point; g(n) represents the actual cost function from the starting point to the current point n; and h(n) represents the cost function from the current point n to the target point.

[0019] Euclidean distance is used to describe the cost of movement between two nodes:

[0020]

[0021] In the formula, Representative node The x-coordinate; Representative node The ordinate;

[0022] S122, Node expansion method: Based on robot kinematics, the geometric constraint of the robot's minimum turning radius under average speed is considered, that is, the radius of curvature of each node on the trajectory is greater than the minimum turning radius of the vehicle. An optimized partial node expansion method is used to optimize the node expansion method, that is, more than six expansion nodes are selected.

[0023] S123. Cost function optimization: Based on step S121, the robot heading cost function t(n) and the ducted fan motion cost b(n) are introduced:

[0024]

[0025] in, Indicates the initial node. Indicates the current node, Indicates the target node; For heading and orientation cost coefficients, For the robot at the current node The heading angle on the top, For the robot at the current node The heading angle on the parent node; The coefficient for the motion cost of the ducted fan. The motion cost when the ducted fan causes the robot's posture to turn 180°;

[0026] The improved cost function is:

[0027]

[0028] Further, in step S200, the following are included:

[0029] S210. Establish a six-degree-of-freedom rigid body kinematics and dynamics model for a space intelligent flight robot under microgravity conditions, including translational mechanics:

[0030]

[0031] in Indicates the position of the robot's center of gravity. Represents the robot's speed. Represents total thrust. Represents the total mass. Represents external disturbance forces;

[0032] The rotational kinematics and dynamics equations are expressed in quaternions as follows:

[0033]

[0034] in Represents a quaternion of postures. Represents the angular velocity of the machine body. Represents the robot's inertia matrix. Represents the total control torque. Represents the disturbance torque. The transformation matrix representing angular velocity to quaternions;

[0035] S220, The PC model predictive control method seeks a set of control inputs over N future time steps. This minimizes the error between the predicted state and the reference state; the optimization problem to be solved is as follows:

[0036]

[0037] in, The optimization variables after using model predictive control methods, To predict the size of the time domain, Here is the state error weight matrix. The weight matrix is ​​used to control the rate of change. This is the weight matrix for the terminal state error. for Predicted trajectory points at time [time] for Reference trajectory points at any given time for The rate of change of the control input at any given time, Predict trajectory points for the terminal at any given time. These are the reference trajectory points at the terminal moment;

[0038] S230: First, it constructs intelligent behavior models for astronauts and floating objects, and uses deep learning to recognize the movements of the astronaut's torso and limbs. Second, based on international human-machine safety standards, it determines the astronaut's future movement range and generates risk areas for robot movement based on the astronaut's movement direction, speed, and acceleration level. Third, based on the human positioning and human intention recognition results provided by the navigation system, it uses the Kalman filter algorithm to predict the position information of human joints and generates obstacle areas in three-dimensional space in real time based on the prediction results.

[0039] Furthermore, step S200 also includes,

[0040] S240. In addition to the astronauts, the three-dimensional map built by the perception system is based on information about spatial obstacles within the confined space. Divided into three areas:

[0041]

[0042] Where O, F, and U represent barrier-free space, obstacle space, unknown space, or risk space, respectively.

[0043] Furthermore, step S200 also includes,

[0044] S250. In the process of solving the objective function, the following constraints are applied based on the robot's mobility and task requirements:

[0045] Unknown obstacle avoidance constraint: When obstacles that cannot be identified at a distance, such as astronaut limb movements, suddenly appear in the motion path, the trajectory tracking algorithm realizes emergency obstacle avoidance;

[0046] Attitude angle and angular velocity constraints: To ensure the normal operation of the robot navigation system, the attitude angle and angular velocity during trajectory tracking are limited to enable it to complete smooth motion.

[0047]

[0048] in,, These are the angle and angular acceleration when the robot turns; , These are the minimum and maximum angles when the robot turns; , These are the minimum and maximum angular accelerations of the robot when it turns;

[0049] Acceleration constraints:

[0050]

[0051] in, , , These represent the robot's accelerations along the x, y, and z axes, respectively. , These represent the robot's minimum and maximum accelerations along the x-axis, respectively. , These represent the robot's minimum and maximum accelerations along the y-axis, respectively. , These represent the robot's minimum and maximum accelerations along the z-axis, respectively.

[0052] The above two constraints can be combined and simplified as follows:

[0053]

[0054] in, For the state of the system, To control the input, For interference.

[0055] The flight robot perception uncertainty motion planning system has program modules corresponding to the above steps, and executes the steps in the flight robot perception uncertainty motion planning method described above when running.

[0056] A computer-readable storage medium storing a computer program configured to implement, when invoked by a processor, steps of a method for perceiving uncertain motion planning in a flying robot.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] This invention discloses a method and system for motion planning of a flying robot with uncertain perception. It employs an optimized partial node expansion method to ensure the final path meets mission requirements while conserving system computational resources and improving path search efficiency. Furthermore, cost function optimization considers both the Euclidean distance cost of the trajectory and the robot's orientation towards the astronaut for easy viewing of screen information. It also encourages the robot to utilize flywheels for attitude adjustment as much as possible to reduce noise from the ducted fan. By transforming uncertainties in perception and modeling into Tube constraints, Tube-MPC is used for backend trajectory optimization to obtain a robust motion trajectory that meets the relevant constraints. Attached Figure Description

[0059] Figure 1 This is a three-dimensional raster map of the cabin environment in an embodiment of the present invention;

[0060] Figure 2 This is a flowchart of the traditional A* algorithm in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the trajectory after node expansion considering kinematic constraints in an embodiment of the present invention;

[0062] Figure 4 This is a flowchart of a method for perceiving uncertain motion planning for a flying robot, as described in an embodiment of the present invention. Detailed Implementation

[0063] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0064] Specific Implementation Plan 1: Combining Figures 1 to 4 As shown, the present invention provides a method for planning uncertain motion for a flying robot, comprising the following steps:

[0065] S100: The robot performs global path planning using an improved A* algorithm based on a 3D grid map.

[0066] Combination Figures 1 to 3 As shown, the robot performs global path planning using an improved A* algorithm based on a 3D grid map. The specific steps are as follows:

[0067] The S110, an onboard core processor, first generates positioning data based on the binocular images acquired by the forward-looking obstacle avoidance camera. Then, it fuses this positioning data with the position data from the rear-looking navigation camera and laser ranging data to obtain higher-precision robot position data. Based on this robot position data, combined with the forward-looking depth point cloud image and the binocular fisheye stereo vision image from the rear-looking navigation camera, it uses the computationally efficient TSDF (Truncation Sign Function) method to map the surrounding environment. Figure 1 );

[0068] S120, the robot uses an improved A* algorithm to perform global path planning based on a 3D grid map, which is an improvement over the traditional A* algorithm ( Figure 2 To address the shortcomings in robot path planning, this invention optimizes and improves the traditional A* algorithm in two aspects: node expansion method and evaluation function optimization. The specific steps are as follows:

[0069] The evaluation function f(n) of the S121 and A* algorithms is expressed as:

[0070]

[0071] In the formula, f(n) represents the cost function from the starting point (via the current point n) to the target point; g(n) represents the actual cost function from the starting point to the current point n; and h(n) represents the cost function from the current point n to the target point.

[0072] Euclidean distance is typically used to describe the cost of movement between two nodes.

[0073]

[0074] In the formula, Representative node The x-coordinate; Representative node The ordinate;

[0075] The following improvements were made:

[0076] S122. Node Expansion Method: Traditional A* algorithms typically employ omnidirectional node expansion for complete systems. However, when the robot requires the forward-looking obstacle avoidance camera to be aligned with the motion velocity direction, it becomes a nonholonomic, constrained system. Therefore, this invention, based on robot kinematics, considers the geometric constraint of the robot's minimum turning radius at average velocity and uses an optimized partial node expansion method to optimize the node expansion approach; for example... Figure 3 Taking the six trajectories shown as an example, the radius of curvature at each point on the trajectory is greater than the minimum turning radius of the vehicle. This ensures the feasibility of the final path while saving system computing resources and improving path search efficiency. In addition, although expanding to six nodes has basically met the requirements of path search, if the system's computing power is fast enough, more expansion nodes can be selected to obtain more accurate planning results.

[0077] S123. Cost Function Optimization: This section compares the cost functions of the traditional A* algorithm with g(n) (the actual cost function from the starting point to the current point n) and h(n) (the estimated cost function from the current point n to the target point). It introduces t(n) (the robot's heading cost function) and b(n) (the ducted fan motion cost function). For the former, we consider that the robot should face the astronaut's direction to facilitate the astronaut's viewing of screen information. For the latter, we aim to use flywheels as much as possible to adjust the robot's attitude, thereby reducing the noise from the ducted fan.

[0078]

[0079] in, Indicates the initial node. Indicates the current node, Indicates the target node; For heading and orientation cost coefficients, Let the robot's heading angle be the current node. This is the robot's heading angle on the parent node of the current node; The coefficient for the motion cost of the ducted fan. The motion cost when the ducted fan causes the robot's posture to turn 180°;

[0080] Therefore, the new cost function is:

[0081]

[0082] This method enables the path to be more suitable for the movement of intelligent space robots in complex, unstructured microgravity environments.

[0083] S200: The robot performs local path planning based on the trajectory points generated by global path planning, the motion information of the astronauts and floating objects inside the cabin, kinematic and dynamic constraints, and Tube constraints.

[0084] Combination Figure 4 As shown, the robot performs local path planning based on the trajectory points generated by global path planning and the motion information of the astronauts and floating objects inside the cabin. Using Model Predictive Control (MPC), the actual safety conditions are modeled as inequality constraints in an optimization problem, and the deviation between the actual trajectory and the front-end trajectory, as well as control energy consumption, are modeled as objective functions for evaluating the motion trajectory. By solving for the optimal value of the objective function in the prediction time domain and its corresponding motion trajectory, safe, feasible, and optimal motion trajectory planning is achieved. The specific steps are as follows:

[0085] S210. Establish a 6-DOF rigid body kinematics and dynamics model for a space intelligent flight robot under microgravity conditions. For translational mechanics, we have:

[0086]

[0087] in Indicates the position of the robot's center of gravity. Represents the robot's speed. Represents total thrust. Represents the total mass. Represents external disturbance forces;

[0088] The rotational kinematics and dynamics equations are expressed in quaternions as follows:

[0089]

[0090] in Represents a quaternion of postures. Represents the angular velocity of the machine body. Represents the robot's inertia matrix. Represents the total control torque. Represents the disturbance torque. The transformation matrix representing angular velocity to quaternions;

[0091] S220 and MPC search for a set of control inputs in the next N time steps. The goal is to minimize the error between the predicted state and the reference state, while also keeping the rate of change of the control input as small as possible, thus minimizing the error of the final state. The optimization problem to be solved is as follows:

[0092]

[0093] in, For the optimization variables after using the Model Predictive Control (MPC) method, To predict the size of the time domain, Here is the state error weight matrix. The weight matrix is ​​used to control the rate of change. This is the weight matrix for the terminal state error. for Predicted trajectory points at time. for Reference trajectory points at any given time for The rate of change of the control input at any given time, Predict trajectory points for the terminal at any given time. These are the reference trajectory points at the terminal moment;

[0094] S230: First, it constructs intelligent behavior models for astronauts and floating objects, and uses deep learning to recognize the movements of the astronaut's torso and limbs. Second, based on international human-machine safety standards, it determines the astronaut's future movement range and generates risk areas for robot movement based on the astronaut's movement direction, speed, and acceleration level. Third, based on the human positioning and human intention recognition results provided by the navigation system, it uses the Kalman filter algorithm to predict the position information of human joints and generates obstacle areas in three-dimensional space in real time based on the prediction results.

[0095] S240. In addition to astronauts, the cabin walls and internal equipment will also constitute spatial obstacles during the robot's movement; based on obstacle information, the three-dimensional map built by the perception system can be... Divided into three areas:

[0096]

[0097] Where O, F, and U represent unobstructed space, obstacle space, unknown space, or risk space, respectively. Considering that the robot is affected by disturbances during its movement, there is a deviation between the actual trajectory and the expected trajectory. It is necessary to design a control algorithm with good anti-interference effect to achieve trajectory tracking control. At the same time, the control quantity output by the trajectory tracking algorithm must also meet various constraints determined by the robot's own physical properties and the passable space.

[0098] S250. To ensure that the control commands generated by the model predictive control can be executed by the robot, the following constraints need to be applied during the solution of the objective function, based on the robot's maneuverability and task requirements:

[0099] Attitude Angle and Angular Velocity Constraints: Since the robot's attitude angle and angular velocity have a significant impact on the sensor's sensing range, it is necessary to limit the attitude angle and angular velocity during trajectory tracking to ensure smooth movement and guarantee the normal operation of its navigation system.

[0100]

[0101] in, , These are the angle and angular acceleration when the robot turns; , These are the minimum and maximum angles when the robot turns; , These are the minimum and maximum angular accelerations of the robot when it turns;

[0102] Acceleration Constraints: Due to the influence of thrust and drag during robot motion, the normal and longitudinal accelerations that a robot can generate are limited. When the output of the MPC algorithm exceeds the limit, the actuator will be unable to execute control commands. Therefore, it is necessary to limit the acceleration output.

[0103]

[0104] in, , , These represent the robot's accelerations along the x, y, and z axes, respectively. , These represent the robot's minimum and maximum accelerations along the x-axis, respectively. , These represent the robot's minimum and maximum accelerations along the y-axis, respectively. , These represent the robot's minimum and maximum accelerations along the z-axis, respectively.

[0105] The two constraints mentioned above can be combined and simplified into the following form:

[0106]

[0107] in, For the state of the system, To control the input, For interference;

[0108] Specific Implementation Scheme 2: The present invention provides a flight robot uncertain motion planning system, which has a program module corresponding to the above steps, and executes the steps in the above flight robot uncertain motion planning method when running.

[0109] The other combinations and connections in this implementation scheme are the same as in Specific Implementation Scheme 1.

[0110] Specific Implementation Scheme 3: The present invention provides a computer-readable storage medium storing a computer program configured to implement, when called by a processor, the steps of a method for perceiving uncertain motion planning of a flying robot.

[0111] The other combinations and connections in this implementation scheme are the same as in Specific Implementation Scheme 1.

[0112] Simulation Experiment

[0113] The method of this invention was applied to the aforementioned motion planning task, and 100 motion planning experiments were repeated. The mean and standard deviation of the 100 experimental data are listed in Table 1. As can be seen from Table 1, the perceptual uncertainty autonomous motion planning algorithm proposed in this invention has low time cost and high energy efficiency. The method of this invention can effectively complete the motion planning task of space intelligent flight robots in microgravity environments and has high engineering application value.

[0114] Table 1. Statistical Analysis of Experimental Data

[0115]

[0116] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for planning uncertain motion for a flying robot, characterized in that, Includes the following steps: S100. The robot performs global path planning using an improved A* algorithm based on a 3D grid map. The improved A* algorithm is based on the traditional A* algorithm, but with node expansion and evaluation function optimization, to obtain the trajectory points generated by the global path planning. include, The S110 airborne core processor first generates positioning data based on the binocular images collected by the forward-looking obstacle avoidance camera. Then, it fuses the positioning data with the position data collected by the rear-looking navigation camera and the laser ranging data to obtain higher-precision robot position data. Based on the robot position data, combined with the forward-looking depth point cloud image and the binocular fisheye stereo vision image of the rear-looking navigation camera, the TSDF method is used to map the surrounding environment. S120. The robot uses an improved A* algorithm to perform global path planning based on a 3D grid map, including node expansion and evaluation function optimization. Specifically, including, The evaluation function f(n) of the S121 and A* algorithms is expressed as: In the formula, f(n) represents the cost function from the starting point to the target point; g(n) represents the actual cost function from the starting point to the current point n; and h(n) represents the cost function from the current point n to the target point. Euclidean distance is used to describe the cost of movement between two nodes: In the formula, Representative node The x-coordinate; Representative node The ordinate; S122, Node Expansion Method: Based on robot kinematics, considering the geometric constraint of the robot's minimum turning radius at average speed, an optimized partial node expansion method is used to optimize the node expansion method, i.e., selecting more than six expansion nodes. S123. Cost function optimization: Based on step S121, the robot heading cost function t(n) and the ducted fan motion cost b(n) are introduced: in, Indicates the initial node. Indicates the current node, Indicates the target node; For heading and orientation cost coefficients, For the robot at the current node The heading angle on the top, For the robot at the current node The heading angle on the parent node; The coefficient for the motion cost of the ducted fan. The motion cost when the ducted fan causes the robot's posture to turn 180°; The improved cost function is: S200: Based on the trajectory points generated by the global path planning in step S100, the robot performs local path planning using the motion information of the astronauts and floating objects inside the cabin, kinematic and dynamic constraints, and Tube constraints. Through model predictive control, the actual safety conditions are modeled as inequality constraints of an optimization problem, the uncertainties of perception and model are transformed into Tube constraints, and the goals of accuracy and energy consumption are modeled as objective functions for evaluating the motion trajectory. By solving for the optimal value of the objective function in the prediction time domain and its corresponding motion trajectory, the robot's motion trajectory planning is achieved. include, S210. Establish a six-degree-of-freedom rigid body kinematics and dynamics model for a space intelligent flight robot under microgravity conditions, including translational mechanics: in Indicates the position of the robot's center of gravity. Represents the robot's speed. Represents total thrust. Represents the total mass. Represents external disturbance forces; The rotational kinematics and dynamics equations are expressed in quaternions as follows: in Represents a quaternion of postures. Represents the angular velocity of the machine body. Represents the robot's inertia matrix. Represents the total control torque. Represents the disturbance torque. The transformation matrix representing angular velocity to quaternions; S220, The PC model predictive control method seeks a set of control inputs over N future time steps. This minimizes the error between the predicted state and the reference state; the optimization problem to be solved is as follows: in, The optimization variables after using model predictive control methods, To predict the size of the time domain, Here is the state error weight matrix. The weight matrix is ​​used to control the rate of change. This is the weight matrix for the terminal state error. for Predicted trajectory points at time [time] for Reference trajectory points at any given time for The rate of change of the control input at any given time, Predict trajectory points for the terminal at any given time. These are the reference trajectory points at the terminal moment; S230: First, it constructs intelligent behavior models for astronauts and floating objects, and uses deep learning to recognize the movements of the astronaut's torso and limbs. Second, based on international human-machine safety standards, it determines the astronaut's future movement range and generates risk areas for robot movement based on the astronaut's movement direction, speed, and acceleration level. Third, based on the human positioning and human intention recognition results provided by the navigation system, it uses the Kalman filter algorithm to predict the position information of human joints and generates obstacle areas in three-dimensional space in real time based on the prediction results.

2. The method for planning uncertain motion of a flying robot according to claim 1, characterized in that: Step S200 also includes, S240. In addition to the astronauts, the three-dimensional map built by the perception system is based on information about spatial obstacles within the confined space. Divided into three areas: Where O, F, and U represent barrier-free space, obstacle space, unknown space, or risk space, respectively.

3. The method for planning uncertain motion of a flying robot according to claim 2, characterized in that: Step S200 also includes, S250. In the process of solving the objective function, the following constraints are applied based on the robot's mobility and task requirements: Unknown obstacle avoidance constraint: When obstacles that cannot be identified at a distance, such as astronaut limb movements, suddenly appear in the motion path, the trajectory tracking algorithm realizes emergency obstacle avoidance; Attitude angle and angular velocity constraints: To ensure the normal operation of the robot navigation system, the attitude angle and angular velocity during trajectory tracking are limited to enable it to complete smooth motion. in, , These are the angle and angular acceleration when the robot turns; , These are the minimum and maximum angles when the robot turns; , These are the minimum and maximum angular accelerations of the robot when it turns; Acceleration constraints: in, , , These represent the robot's accelerations along the x, y, and z axes, respectively. , These represent the robot's minimum and maximum accelerations along the x-axis, respectively. , These represent the robot's minimum and maximum accelerations along the y-axis, respectively. , These represent the robot's minimum and maximum accelerations along the z-axis, respectively. The above two constraints can be combined and simplified as follows: in, For the state of the system, To control the input, For interference.

4. A motion planning system for a flying robot that perceives uncertainties, characterized in that: The system has a program module corresponding to the steps described in any one of claims 1-3 above, and executes the steps in the above-described method for perceiving uncertain motion planning of flying robots when it is run.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program configured to implement, when invoked by a processor, the steps of the flight robot's uncertain motion planning method as described in any one of claims 1-3.

Images