Obstacle avoidance re-planning method and system considering dynamic constraints and operating limit conditions

Abstract

The application belongs to the technical field of intelligent agent path planning, and discloses an obstacle avoidance re-planning method and system considering kinetic constraints and operation limit conditions, which comprises the following steps: constructing an intelligent agent and obstacle interaction model considering the safety distance of obstacles; setting a dynamic target area in front of the intelligent agent moving with the intelligent agent; obtaining an initial optimized trajectory and an initial optimized motion state by using a path planning method; abstracting the intelligent agent as an intelligent agent point mass model, drawing two straight lines parallel to the axis and the axis with the intelligent agent point mass model as the starting point, and the two straight lines intersecting with the initial optimized trajectory at two points, taking the point closest to the target point among the two points as the final target point; constructing a predictive control model; and solving the predictive control model to obtain continuous final optimized operation state and final optimized trajectory. The application can optimize the existing path planning and improve the obstacle avoidance success rate.

Description

Technical Field

[0001] This invention belongs to the technical field of obstacle avoidance in robot motion, and more specifically, relates to an obstacle avoidance replanning method and system that considers dynamic constraints and operational limitations. Background Technology

[0002] Obstacle avoidance planning methods are crucial for enabling autonomous navigation of intelligent agents. Current obstacle avoidance planning methods typically target specific global path maps for goal-oriented motion planning. Designing excellent local obstacle avoidance methods requires consideration not only of obstacle detection and optimal path planning, but also the agent's motion control performance and robustness. Furthermore, to ensure the agent can move stably in complex environments and complete its mission objectives, dynamic constraints and motion condition limitations must be considered. Commonly used obstacle avoidance planning methods include Artificial Potential Field (APF), Dynamic Window (DWA), and Vector Field Histogram (VFH), all of which demonstrate excellent performance in local obstacle avoidance planning. However, existing obstacle avoidance planning methods still have some problems in specific application scenarios: 1) The algorithm design only considers graph-level motion planning and does not update the trajectory according to the actual dynamic constraints of the agent; 2) In open scenarios with many obstacles, the agent's motion is often subject to many motion condition limitations. Therefore, research on new local obstacle avoidance planning strategies that consider dynamic constraints and motion limitations is very meaningful. Summary of the Invention

[0003] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides an obstacle avoidance replanning method and system that takes into account dynamic constraints and operational limitations, which can optimize the existing path planning and improve the obstacle avoidance success rate.

[0004] To achieve the above objectives, according to one aspect of the present invention, an obstacle avoidance replanning method considering dynamic constraints and motion limitations is provided. The method includes: S1: constructing an interaction model between an agent and an obstacle considering the safe distance of the obstacle; S2: setting a dynamic target area in front of the agent that moves with the agent; S3: using a path planning method to continuously select target points in the dynamic target area to optimize the preset trajectory of the agent to obtain an initial optimized trajectory and an initial optimized motion state; S4: abstracting the agent into an agent point mass model, drawing two straight lines parallel to the X-axis and Y-axis from the agent point mass model as the starting point, the two straight lines intersecting the initial optimized trajectory at two points, and taking the point closest to the target point as the final target point; S5: constructing a predictive control model using the final motion state to be determined, the initial optimized motion state, the control input function, and the penalty function; S6: transforming the predictive control model and solving it to obtain discrete final optimized running states and final optimized trajectories; S7: using curve polynomial fitting to obtain continuous final optimized running states and final optimized trajectories.

[0005] Preferably, in step S1, the interaction model between the agent and the obstacle is as follows:

[0006]

[0007] in, To avoid collision conditions, when When the conditions are met, the intelligent agent can operate safely. x and y represent the position coordinates of the agent, respectively. and d represents the coordinates of the obstacle's center. s d represents the radar scanning range. ero To account for the obstacle expansion radius of the safety distance, d ero =d ero1 +d ero2 d ero2 To preset a safe distance, d ero1 Let be the radius of the obstacle's expansion.

[0008] Preferably, the formula for the expansion radius of the obstacle is:

[0009]

[0010] Where, r rotation The radius of rotation of the intelligent agent itself. This refers to the potential range of interference that may occur during the operation of the intelligent agent. For the maximum value among them, This represents the maximum distance from the boundary of the agent's interference range to the obstacle.

[0011] Preferably, in step S2, the dynamic target region is formed by expanding outwards from the center of the dynamic target point region, wherein the forward movement strategy of the center of the dynamic target point region is designed as follows:

[0012]

[0013] in, Let Γ be the dynamic target point region at time t, and let V be the adjustment factor for the movement of the dynamic target point region. l for The velocity along the original reference trajectory, T is the sampling period, d is the initial distance to the dynamic target point region associated with the obstacle, τ is the yaw correction parameter, and υ is the obstacle size correction parameter, τ, υ∈R + ψ(t) is the angle between the direction the agent's head is facing and the tangent direction of the dynamic target point on the original trajectory at each moment. The angle of the sensor for detecting obstacles.

[0014] Preferably, the expression for the dynamic target region is:

[0015]

[0016] in, and Let κ ∈ R represent the upper and lower bounds of the dynamic target point region, respectively. + D(x) is the dynamic target point region.

[0017] Preferably, in step S4, the agent point quality model is:

[0018]

[0019] |μ(t)|<ug

[0020] Where ξ(t) represents the motion state of the agent at time t. Let x be the velocity of the agent in the x-axis direction. Let θ be the velocity of the agent along the y-axis, θ be the heading angle, Y be the ordinate of the agent's position at time t, and X be the abscissa of the agent's position at time t. Let ξ(t) be the derivative of ξ(t), μ(t) be the input at time t, g be the gravitational acceleration, and u be the friction coefficient.

[0021] Preferably, in step S5, the predictive control model is:

[0022]

[0023] stU min ≤U t ≤Umax

[0024] Where, N p To predict the time domain, U i To control the set of input points, J obs，i Let Q be the penalty function applied at time i, and let U be the weighting coefficient. min and U max These are its maximum and minimum values, ξ. ref (t+i|t) represents the initial optimized motion state parameters. U t To control the input sequence, To initially optimize the agent's velocity along the y-axis during motion, To initially optimize the agent's velocity along the x-axis in its motion state, θ ref To initially optimize the agent's heading angle in motion, Y ref To initially optimize the agent's position ordinate in the motion state, X ref The x-coordinate of the agent's position in the initial optimization motion state;

[0025] The penalty function is:

[0026]

[0027] Among them, S obs V represents the weighting coefficient. l Let x be the magnitude of the longitudinal velocity of the intelligent agent's chassis. i y i (x0, y0) represents the position coordinates of the obstacle point cloud in the agent's coordinate system, and (x0, y0) represents the coordinates of the agent's centroid. It is a small positive number.

[0028] Preferably, in step S6, the predictive control model is transformed to obtain a standard quadratic programming form, and then solved using the built-in programming solver of MATLAB.

[0029] Preferably, the obstacles need to be processed as follows: based on the obstacle point cloud information provided by the lidar and infrared sensors; the obstacle point clouds with a distance less than a preset distance are merged into large obstacles using a clustering algorithm; and the obstacle point clouds with a distance greater than the preset distance are regarded as discrete small obstacles.

[0030] Preferably, the path planning method in step S3 is DWA, VFH, or APF.

[0031] This application, in another aspect, provides a local replanning system considering dynamic constraints and motion limitations. The system includes: a first construction module for constructing an agent and obstacle interaction model considering obstacle safety distances; a setting module for setting a dynamic target area in front of the agent that moves with it; a preliminary planning module for using path planning to continuously select target points in the dynamic target area to optimize the agent's preset trajectory and obtain an initial optimized trajectory and initial optimized motion state; a target point acquisition module for abstracting the agent into an agent point mass model, drawing two straight lines parallel to the x-axis and y-axis from the agent point mass model, intersecting the initial optimized trajectory at two points, and taking the point closest to the target point as the final target point; a second construction module for constructing a predictive control model using the final motion state to be determined, the initial optimized motion state, a control input function, and a penalty function; a solution module for transforming the predictive control model and solving it to obtain discrete final optimized operating states and final optimized trajectories; and a fitting module for using curve polynomial fitting to obtain continuous final optimized operating states and final optimized trajectories.

[0032] In summary, compared with the prior art, the obstacle avoidance replanning method and system considering dynamic constraints and operational limitations provided by the present invention have the following advantages:

[0033] This application achieves initial optimization by setting a dynamic target area in which the agent moves along a preset trajectory after obstacle avoidance. Furthermore, by abstracting the agent into a point mass model and designing a predictive control model that includes the final motion state, the initial optimized trajectory and initial motion state are further optimized to improve the obstacle avoidance success rate.

[0034] This application incorporates a penalty function into the predictive control model to control the distance deviation between obstacles, ensuring that the closer an obstacle is to another obstacle, the larger the function value, thereby correcting the path.

[0035] By setting up a dynamic target region and selecting a target point within it, the agent can return to the original path after avoiding obstacles. This improves the utilization rate of the original reference path compared to traditional methods and significantly reduces the computational burden of real-time planning. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the steps of the obstacle avoidance replanning method considering dynamic constraints and motion limitations in an embodiment of this application.

[0037] Figure 2 This is a flowchart of the obstacle avoidance replanning method considering dynamic constraints and motion limitations in the embodiments of this application;

[0038] Figure 3A This is a schematic diagram of the obstacle expansion radius calculation method according to an embodiment of this application;

[0039] Figure 3B This is a schematic diagram illustrating the principle of the obstacle expansion radius calculation method in this application embodiment;

[0040] Figure 4 This is a schematic diagram of the dynamic target point region in an embodiment of this application;

[0041] Figure 5 This is the method for finding the final target point in the embodiments of this application. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0043] This invention provides an obstacle avoidance reprogramming method that considers dynamic constraints and motion limitations, such as... Figure 1 and Figure 2 As shown, the method includes the following steps S1 to S7.

[0044] S1: Construct an agent-obstacle interaction model that considers safe distance from obstacles.

[0045] Specifically, the interaction model between the agent and the obstacle is as follows:

[0046]

[0047] in, To avoid collision conditions, when When the conditions are met, the intelligent agent can operate safely. x and y represent the position coordinates of the agent, respectively. and d represents the coordinates of the obstacle's center. s d represents the radar scanning range. ero To account for the obstacle expansion radius considering the safety distance, such as Figure 3A and 3B As shown, d ero =d ero1 +d ero2 d ero2To establish a preset safe distance and avoid collisions between the agent and obstacles in uncertain environments, this preset safe distance can offset the agent's state shift caused by radar sensing positioning accuracy errors. ero1 Let be the radius of the obstacle's expansion.

[0048] More preferably, the formula for the expansion radius of the obstacle is:

[0049]

[0050] Where, r rotation The radius of rotation of the intelligent agent itself. This refers to the potential range of interference that may occur during the operation of the intelligent agent. For the maximum value among them, This represents the maximum distance from the boundary of the agent's interference range to the obstacle.

[0051] S2: Set a dynamic target area in front of the agent that moves with the agent.

[0052] When obstacle avoidance mode is activated, a designated area in front of the current agent's location is set as the dynamic target area for the agent's obstacle avoidance algorithm, such as... Figure 4 As shown, the dynamic target area will move at a constant speed along the original preset trajectory.

[0053] More preferably, the dynamic target region is formed by expanding outwards from the center of the dynamic target point region, wherein the forward movement strategy of the center of the dynamic target point region is designed as follows:

[0054]

[0055] in, Let Γ be the dynamic target point region at time t, and let V be the adjustment factor for the movement of the dynamic target point region. l Let d be the velocity of the dynamic target point region moving along the reference path, T be the sampling period, d be the initial distance of the dynamic target point region associated with the obstacle, τ be the yaw correction parameter, and υ be the obstacle size correction parameter, where τ, υ ∈ R. + ψ(t) is the angle between the direction the agent's head is facing and the tangent direction of the dynamic target point on the original trajectory at each moment. The angle of the sensor for detecting obstacles.

[0056] The expression for the dynamic target region is:

[0057]

[0058] in, and Let κ ∈ R represent the upper and lower bounds of the dynamic target point region, respectively.+ D(x) is the dynamic target point region.

[0059] A further preferred solution also includes the following treatment of obstacles:

[0060] Based on obstacle point cloud information provided by lidar and infrared sensors, the location of obstacles in the environment in which the intelligent agent is located is represented.

[0061] The point clouds of obstacles with a distance less than a preset distance are merged into large obstacles using a clustering algorithm;

[0062] For obstacle point clouds that are more than a preset distance away, they are considered as discrete small obstacles.

[0063] After processing the obstacle as described above, calculate the obstacle's expansion radius.

[0064] S3: The path planning method is used to continuously select target points in the dynamic target area to optimize the preset trajectory of the agent and obtain the initial optimized trajectory and initial optimized motion state.

[0065] After obtaining the dynamic target point region, a suitable target point is selected as the initial replanning guide. The planning method can be DWA, VFN or APF, which enables the agent to return to the original path after obstacle avoidance, thereby greatly reducing the system's computational load and improving the efficiency of obstacle avoidance replanning.

[0066] S4: Abstract the agent into an agent point quality model. Starting from the agent point quality model, draw two straight lines parallel to the X-axis and Y-axis. The two straight lines intersect the initial optimized trajectory at two points. The point that is closest to the target point among the two points is taken as the final target point.

[0067] The agent point quality model is designed as follows:

[0068]

[0069] If we consider dynamic constraints, the limiting conditions can be described in the following form:

[0070] |a y |<ug

[0071] Combining the above two equations, the agent point quality model can be simplified to the following form:

[0072]

[0073] |μ(t)|<ug

[0074] Where ξ(t) represents the motion state of the agent at time t. Let x be the velocity of the agent in the x-axis direction. Let θ be the velocity of the agent along the y-axis, θ be the heading angle, Y be the ordinate of the agent's position at time t, and X be the abscissa of the agent's position at time t. Let ξ(t) be the derivative of ξ(t), μ(t) be the input at time t, g be the gravitational acceleration, and u be the friction coefficient.

[0075] An agent can find the point closest to the target point as the final target point by drawing two straight lines parallel to the x and y axes in the global coordinate system and intersecting them with the initial optimized trajectory. Figure 5 As shown, this avoids the problem of frequent turning.

[0076] S5: Construct a predictive control model using the final motion state to be determined, the initial optimized motion state, the control input function, and the penalty function.

[0077] The predictive control model is as follows:

[0078]

[0079] stU min ≤U t ≤U max

[0080] Where ξ(t+i|t)-ξ ref (t+i|t) represents the deviation between the trajectory to be predicted and the initial optimized trajectory, N p To predict the time domain, U i To control the set of input points, J obs，i Let Q be the penalty function applied at time i, and let U be the weighting coefficient. min and U max For its maximum and minimum values, ξ ref (t+i|t) represents the initial optimized motion state parameters. U t To control the input sequence, To initially optimize the agent's velocity along the y-axis during motion, To initially optimize the agent's velocity along the x-axis in its motion state, θ ref To initially optimize the agent's heading angle in motion, Y ref To initially optimize the agent's position ordinate in the motion state, X ref The x-coordinate of the agent's position in the initial optimization motion state;

[0081] The penalty function is:

[0082]

[0083] Among them, S obsV represents the weighting coefficient. l Let x be the magnitude of the longitudinal velocity of the intelligent agent's chassis. i y i (x0, y0) represents the position coordinates of the obstacle point cloud in the agent's coordinate system, and (x0, y0) represents the coordinates of the agent's centroid. It is a small positive number. The design of the penalty function requires calculating the distance deviation between the obstacle and the final target point, and adjusting the function value accordingly to ensure that the closer the obstacle is to the target point, the larger the function value.

[0084] Typically, since obstacle avoidance planners do not have high real-time requirements, the agent's point quality model is sufficient to meet their basic needs. However, to improve solution accuracy, a higher-level nonlinear predictive control algorithm can be used to meet the requirements of obstacle avoidance replanning. Furthermore, using a nonlinear objective function also facilitates the design of subsequent penalty functions. Therefore, linearization of the predictive control model is no longer necessary; the solution can be directly based on the nonlinear model.

[0085] S6: After deforming the predictive control model, solve it to obtain the discrete final optimal operating state and final optimal trajectory.

[0086] To facilitate the solution, the predictive control model needs to be transformed before solving. The specific process is as follows.

[0087] The state-space expression of the predictive control model is:

[0088]

[0089] Among them, ξ(k|t)=[ξ(k) μ(k-1)] T For the newly constructed state vector, and These are the parameter matrices of the discrete state equations obtained from the agent's point quality model. u(k) is the control input, Δu(k) is its change, and η(k) is the state output.

[0090] Based on the above formula for state prediction, we can obtain:

[0091]

[0092] Where, N p For prediction in the time domain.

[0093] The system output based on the new spatial state equation described above can be calculated as follows:

[0094]

[0095] After rearranging the above equation, the matrix form of the agent's motion state is obtained as follows:

[0096] γ=Ψζ(k)+ΘΔU

[0097] Where γ is the state output matrix, γ = [η(k+1), ..., η(k+N)] p )] T ,

[0098] To enable the agent to perform obstacle avoidance and replanning while tracking a reference trajectory, the performance evaluation function can be set as follows:

[0099] J = (YY) ref ) T Q Q (YY ref )+ΔU T Q Q ΔU+J obs，i =(Ψξ(k)+Θ t ΔU(t)-Y ref ) T Q Q (Ψξ(k)+Θ t ΔU(t)-Y ref )+ΔU T Q Q ΔU+J obs，i =ΔU(t) T (Θ t T Q Q Θ t +Q Q )ΔU(t)+2(E T Q Q Θ t -Y ref Q Q Θ t )ΔU(t)+J obs，i +E T Q Q E+Y ref T Q Q Y ref -2Y ref T Q Q E

[0100] Where E = Ψζ(k), Let Q be the identity matrix and Q be the weight coefficients. The product is the Kroneck product, ΔU is the sequence of control input changes, and Y is the Kroneck product. ref (k)=[η reef (k+1), ..., ηref (k+N p )]T.

[0101] Ignoring the constant term in the evaluation function, the performance evaluation function can be simplified to:

[0102] J = ΔU T (Θ t T Q Q Θ t +Q Q )ΔU+2(E T Q Q Θ t -Y ref Q Q Θ t )ΔU+J obs，i

[0103] Where E = Ψζ(k).

[0104] Therefore, the predictive control model optimization problem can be transformed into a standard quadratic programming problem, which can be solved using MATLAB's built-in programming solver. This yields the virtual wheel train rotation angle input increment during the trajectory replanning process, and then the discrete final optimized trajectory after obstacle avoidance planning can be obtained.

[0105] S7: Curve polynomial fitting is used to obtain the continuous final optimization running state and final optimization trajectory.

[0106] To ensure the continuity of motion, the discrete final optimized trajectory needs to be optimized and fitted. Curve polynomial fitting is a common linearization method. Due to the kinematic constraints of the agent, namely the need for continuous position information and first-order continuous heading angles, and considering computational efficiency, a fourth-order polynomial is preferred for fitting the discrete trajectory reference point, expressed as:

[0107]

[0108] Where {a4, a3, a2, a1, a0} and {b4, b3, b2, b1, b0} are the fitting polynomial parameters to be calculated, and the replanning path pose information Y is obtained through the above calculation. ref and heading angle information θ ref It can be imported into the system as the final optimized trajectory after completing local obstacle avoidance.

[0109] This application also provides a local replanning system considering dynamic constraints and motion limitations. The system includes a first construction module, a setting module, a preliminary planning module, a target point acquisition module, a second construction module, a solution module, and a fitting module, wherein:

[0110] The first building block: used to build an agent and obstacle interaction model that takes into account the safe distance of obstacles;

[0111] Setting module: Used to set a dynamic target area in front of the agent that moves with the agent;

[0112] Preliminary planning module: used to optimize the agent's preset trajectory and preliminary optimized motion state by continuously selecting target points in the dynamic target area using path planning methods;

[0113] Target point acquisition module: used to abstract the agent into an agent point quality model, draw two straight lines parallel to the X-axis and Y-axis with the agent point quality model as the starting point, the two straight lines intersect the initial optimization trajectory at two points, and take the point closest to the target point as the final target point;

[0114] The second building module is used to construct a predictive control model using the final motion state to be determined, the initial optimized motion state, the control input function, and the penalty function.

[0115] Solving module: used to transform the predictive control model and solve it to obtain the discrete final optimal operating state and final optimal trajectory;

[0116] Fitting module: Used to obtain the continuous final optimization running state and final optimization trajectory by using curve polynomial fitting.

[0117] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for obstacle avoidance replanning considering dynamic constraints and motion limitations, characterized in that, The method includes: S1: Construct an agent-obstacle interaction model that considers safe distance from obstacles; S2: Set a dynamic target area in front of the agent that moves with the agent; S3: The path planning method is used to continuously select target points in the dynamic target area to optimize the preset trajectory of the agent and obtain the initial optimized trajectory and initial optimized motion state; S4: Abstract the agent into an agent point quality model. Starting from the agent point quality model, draw two straight lines parallel to the X-axis and Y-axis. The two straight lines intersect the initial optimization trajectory at two points. The point that is closest to the target point among the two points is taken as the final target point. S5: Construct a predictive control model using the final motion state to be determined, the initial optimized motion state, the control input function, and the penalty function; S6: After deforming the predictive control model, solve it to obtain the discrete final optimal operating state and final optimal trajectory; S7: Curve polynomial fitting is used to obtain the continuous final optimization running state and final optimization trajectory.

2. The method according to claim 1, characterized in that, In step S1, the interaction model between the agent and the obstacle is as follows: in, To avoid collision conditions, when When the conditions are met, the intelligent agent can operate safely. , x and y These represent the position coordinates of the agent. and These represent the coordinates of the obstacle's center. Radar scanning range, The radius of the obstacle expansion is taken into account for the safety distance. , To establish a safe distance, Let be the radius of the obstacle's expansion, and the formula for the radius of the obstacle's expansion is: in, The radius of rotation of the intelligent agent itself. This refers to the potential range of interference that may occur during the operation of the intelligent agent. For the maximum value among them, This represents the maximum distance from the boundary of the agent's interference range to the obstacle.

3. The method according to claim 1, characterized in that, In step S2, the dynamic target region is formed by expanding outwards from the center of the dynamic target point region. The forward movement strategy for the center of the dynamic target point region is designed as follows: in, for t The dynamic target point region at any given time. This is an adjustment factor for the dynamic movement of the target point region. for The speed at which it moves forward along the original reference trajectory. The sampling period is The initial distance to the region of dynamic target points associated with obstacles. These are the yaw correction parameters. For obstacle size correction parameters, , The angle between the agent's head orientation and the tangent direction of the dynamic target point on the original trajectory at each moment. The angle of the sensor for detecting obstacles.

4. The method according to claim 3, characterized in that, The expression for the dynamic target region is: in, and These represent the upper and lower bounds of the dynamic target point region, respectively. , It is a dynamic target point region.

5. The method according to claim 1, characterized in that, In step S4, the agent point quality model is as follows: in, Let t represent the motion state of the agent. , Let x be the velocity of the agent in the x-axis direction. Let be the velocity of the agent in the y-axis direction. For heading angle, Let be the ordinate of the agent's position at time t. Let x be the x-coordinate of the agent's position at time t. for The derivative of The input at time t, It is the acceleration due to gravity. is the coefficient of friction.

6. The method according to claim 1, characterized in that, In step S5, the predictive control model is: in, To predict the time domain, To control the set of input points, To adopt time i The penalty function, Q These are the weighting coefficients. and These represent the maximum and minimum values ​​of the control input sequence, respectively. To initially optimize the motion state parameters, , To control the input sequence, To initially optimize the agent's velocity along the y-axis during motion, To initially optimize the agent's velocity along the x-axis during its motion state, To initially optimize the heading angle of the agent in motion, To initially optimize the agent's position in motion, the ordinate is used. The x-coordinate of the agent's position in the initial optimization motion state; The penalty function is: in, These are the weighting coefficients. The magnitude of the longitudinal velocity of the intelligent agent's chassis. The coordinates of the obstacle point cloud in the agent coordinate system. For the coordinates of the intelligent body's center of mass, It is a small positive number.

7. The method according to claim 1, characterized in that, In step S6, the predictive control model is transformed to obtain a standard quadratic programming form, which is then solved using the built-in programming solver in MATLAB.

8. The method according to claim 1, characterized in that, The obstacle needs to be addressed as follows: Based on obstacle point cloud information provided by lidar and infrared sensors; The point clouds of obstacles with a distance less than a preset distance are merged into large obstacles using a clustering algorithm; For obstacle point clouds that are more than a preset distance away, they are considered as discrete small obstacles.

9. The method according to claim 1, characterized in that, The path planning method in step S3 is DWA, VFH, or APF.

10. A local reprogramming system considering dynamic constraints and motion limitations, characterized in that, The system includes: The first building block: used to build an agent and obstacle interaction model that takes into account the safe distance of obstacles; Setting module: Used to set a dynamic target area in front of the agent that moves with the agent; Preliminary planning module: used to optimize the agent's preset trajectory and preliminary optimized motion state by continuously selecting target points in the dynamic target area using path planning methods; Target point acquisition module: used to abstract the agent into an agent point quality model, draw two straight lines parallel to the X-axis and Y-axis with the agent point quality model as the starting point, the two straight lines intersect the initial optimization trajectory at two points, and take the point closest to the target point as the final target point; The second building module is used to construct a predictive control model using the final motion state to be determined, the initial optimized motion state, the control input function, and the penalty function. Solving module: used to transform the predictive control model and solve it to obtain the discrete final optimal operating state and final optimal trajectory; Fitting module: Used to obtain the continuous final optimization running state and final optimization trajectory by using curve polynomial fitting.

Images