Inspection robot obstacle avoidance method based on fusion of artificial potential field method and dynamic window method

Abstract

The invention provides an inspection robot obstacle avoidance method based on fusion of an artificial potential field method and a dynamic window method, which belongs to the technical field of robot path planning and comprises the following steps: firstly, respectively constructing a gravitational potential field function and a repulsive potential field function according to the distance from a robot to a target point and the distance from the robot to an obstacle; a traditional repulsive force potential field function is modified into a sampling interval, negative gradients of the gravitation potential field function and the repulsive force potential field function are solved to obtain a resultant force set borne by the robot, an optimal path is selected from multiple paths according to an evaluation function, and finally the cyclic sampling prediction process is repeated until the robot reaches the end point. According to the method, the local minimum value points can be effectively avoided, the path with the least redundant nodes is selected, and the obstacle avoidance efficiency is improved on the premise of ensuring safety.

Description

TECHNICAL FIELD

[0001] The application relates to an obstacle avoidance method for a patrol robot based on fusion of an artificial potential field method and a dynamic window method, in particular to an obstacle avoidance method for a patrol robot based on fusion of an artificial potential field method and a dynamic window method. TECHNICAL BACKGROUND

[0002] During movement, a mobile robot needs to rely on an algorithm to calculate a safe and effective driving path in real time to avoid obstacles in the environment and reach the target point. Such an algorithm is a robot obstacle avoidance algorithm, i.e. a local path planning algorithm. Local path planning adjusts the robot's path in real time according to a series of steps based on real-time sensing of the external environment and the position and shape of obstacles by the robot's sensors, and moves towards the target point while avoiding collision with obstacles. The method for the robot to perceive the external environment usually includes sensors and data processing algorithms. Common types of sensors include cameras, ultrasonic sensors, laser radars, infrared sensors, etc. The local path planning algorithm is one of the key technologies for realizing autonomous navigation and obstacle avoidance of mobile robots, and is of great significance for safe and effective driving of mobile robots in complex environments.

[0003] The artificial potential field method is a technique for robot navigation and path planning. This method assumes that the robot moves under the action of a virtual force field, constructs a repulsive potential field around obstacles and an attractive potential field around the target point, similar to electromagnetic fields in physics. The attractive force generated by the target point causes the robot to move towards it, while the repulsive force generated by the obstacles pushes the robot away from them, thus forming a potential field in which the robot adjusts its movement direction according to the gradient of the force. However, the artificial potential field method has some drawbacks. First, in a complex environment, there may be a local minimum problem. In the artificial potential field method, the force acting on the robot is calculated based on the relative position of the obstacles and the target position in the environment. When the resultant force acting on the robot is zero, the robot will stop moving and become trapped in a local minimum point. This situation is more pronounced in environments with high density of obstacles or complex shapes. Second, the design of the potential field may cause the robot to behave unstably in high-density obstacle environments, making it difficult to effectively adjust the path and increasing the complexity of navigation. In addition, adjusting the potential field parameters requires experience and poor adaptability, making it difficult to cope with dynamically changing environments.

[0004] Dynamic window approach (DWA) is a technique for mobile robot obstacle avoidance and path planning, which dynamically generates feasible velocity windows by considering the current speed, acceleration and environmental information of the robot. DWA calculates the states that can be reached within a certain time range based on the motion model of the robot, and evaluates the effectiveness and safety of these states to select the best motion command. This method effectively combines kinematic constraints and environmental information, improving the navigation ability of the robot in dynamic environments. However, DWA also has some drawbacks. The effectiveness of DWA is highly dependent on parameter settings, and improper parameter selection may result in poor obstacle avoidance or reduced navigation performance. In addition, in narrow or complex environments, DWA may not be able to find a suitable path in time, affecting the navigation efficiency of the robot. SUMMARY

[0005] The algorithm aims to solve the problem of redundant nodes caused by the repulsive force of obstacles in the obstacle avoidance process of the robot, and the problem of the robot falling into local minimum points under the action of potential field force, and to improve the passability of mobile robots in complex environments and reduce the dependence on parameter settings.

[0006] To achieve the above purpose, the application adopts the following technical scheme:

[0007] An obstacle avoidance method for inspection robots based on the fusion of artificial potential field method and dynamic window method, comprising the following steps:

[0008] S1: determining the current position coordinates X of the robot k , the position coordinates X of the obstacles O and the end position coordinates X g and related parameters.

[0009] S2: modifying the repulsive potential field function to obtain the repulsive potential field function U rep (X k ) set space, and retaining a repulsive potential field function generated by the repulsive potential field function in the set space for reference.

[0010] S3: calculating the negative gradient of the repulsive potential field function U rep (X k ) set space to obtain the repulsive force set F rep (X k ) = {f rep1 (X k ), f rep2 (X K )……f repn (X k )}, which represents the repulsive potential field function U rep (X kThe steepest descent direction of the inspection robot requires the distance ρ(X) from the obstacle to the inspection robot. k Differentiate (X0).

[0011] S4: The attractive potential field U exerted on the robot from the target point is obtained using the formula for the gravitational potential field function. att (X k )

[0012] S5: Through the repulsive potential field function U att (X k Find the attractive force F acting on the negative gradient mobile robot. att (X k Its physical meaning represents the repulsive potential field function U. att (X k The steepest descent direction of the inspection robot requires the distance ρ(X) from the inspection robot to the target point. k X g Find the derivative.

[0013] S6: Calculate the net force F(X) acting on the mobile robot. k ), resultant force F(X) k The calculation method is as follows:

[0014] F(X k ) = F att (X k )+F rep (X k )

[0015] S7: Calculate F(X) k After the net force on point F(X) is applied, the mobile robot moves towards F(X). k The vector direction movement step s is used to obtain the position coordinates X of point k+1. k+1 Calculate X in the same way k+1 The force at a point does not require sampling in subsequent calculations; the value of q is [not specified].

[0016] S8: Continue calculating the force at subsequent points until X is calculated. m The net force F(X) on the point m This yields the local point set X = {X1, X2, ..., X...} m}

[0017] S9: After obtaining the set of m points, the trajectory set for local path planning is obtained. Under the premise of the same repulsive potential field function, the line connecting adjacent points is the path trajectory TRA that the inspection robot will take. It is also represented by a set TRA = {TRA1, TRA2, ..., TRA}. n}express.

[0018] S10: Introducing evaluation function to evaluate different trajectories, the highest score path is the optimal trajectory in the current point range.

[0019] S11: The evaluation function is composed of the following evaluation indexes:

[0020] S12: (1) Heading evaluation function heading(F att (X k ), F rep (X k )), the evaluation index is the angle Δθ between the vector direction of the trajectory endpoint coordinate and the line connecting the current position coordinate point and the target point, the purpose is to avoid the redundant nodes of the inspection robot path caused by the repulsive force of obstacles.

[0021] S13: (2) The steepest descent evaluation function velocity(F att (X k ), F rep (X k )), the purpose is to adjust the path deviation of the inspection robot caused by the influence of other evaluation functions in the evaluation, avoid deviating from the steepest descent direction, the evaluation index is the Euclidean distance Dist(V) of the coordinate position of the trajectory endpoint and the coordinate position of the target point.

[0022] S14: (3) Local minimum avoidance evaluation function minima(F att (X k ), F rep (X k )), the purpose is to avoid falling into the local minimum point through the evaluation function before the mobile robot falls into the local minimum value.

[0023] S15: Normalize the evaluation functions heading(F att (X k ), F rep (X k )) and velocity(F att (X k ), F rep (X k )).

[0024]

[0025] Where i represents the i-th simulated trajectory.

[0026] S16: The mathematical expression of the evaluation function is as follows.

[0027] E​v (F att (X k ),F rep (X k ))=σ[αheading(F att (X k ),F rep (X k ))]+σ[βvelocity(F att (X k ),F rep (X k ))]+σ[γminima(F att (X k ),F rep (X k ))]

[0028] Wherein, α, β, γ are all coefficients of evaluation function.

[0029] S17: compare the final evaluation scores E v (F att (X k ),F rep (X k )) of n trajectories, and the one with the highest final score is the optimal path selected in one cycle.

[0030] S18: calculate the Euclidean distance dist (X k , X g ) between the current point coordinate X k of the mobile robot and the target point coordinate X g , if dist (X k , X g )<s, it is determined that the target point is reached, and the obstacle avoidance process is ended.

[0031] S19: if dist (X k , X g )>s, return to step S1

[0032] The present application has the following advantages due to the above technical solutions:

[0033] This method modifies the repulsive potential field function, transforming the traditional function from a single value into an interval. This allows for the generation of repulsive forces under various potential field functions, which, together with the attractive force from the target point, constitute the resultant force at the current position coordinates. This resultant force is also an interval. By setting a fixed step size, the method calculates the force on the robot after it moves one step in the direction of the resultant force vector, thus generating the robot's predicted trajectory. The trajectories under different repulsive potential field function conditions require evaluation using an evaluation function. This method selects the path with the highest score as the robot's locally optimal path. The evaluation function metrics consist of the angle between the lines connecting the trajectory endpoints, the steepest descent direction, and the force at the next point. These metrics effectively prevent redundant nodes in the robot's path due to obstacle repulsion and allow for early detection and avoidance of local minima, improving obstacle avoidance efficiency. Attached Figure Description

[0034] Appendix Figure 1 The flowchart of this invention

[0035] Appendix Figure 2 To construct a virtual force field diagram

[0036] Appendix Figure 3 Schematic diagram of the predicted path trajectory Detailed Implementation

[0037] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0038] 1. Determine the relevant parameters and calculate the resultant force acting on the mobile robot.

[0039] As attached Figure 1 The diagram shown is the main flowchart of this invention. First, the position coordinates X of the starting point of the obstacle avoidance process are set. s (x, y) and the coordinates of the target point X g (x, y), the robot's initial position coordinates X k (x, y) represents the position coordinates X of the starting point. s (x,y).

[0040] Set initial parameters, and the proportional gain factor k of the repulsive potential field. rep The proportional gain factor k of the attracted potential field att The maximum influence distance ρ0 of the repulsive potential field of the obstacle, and the variable q randomly selected within the interval (0,1), represented by the set q={q1,q2,q3……q n} indicates that n represents the number of samples randomly selected.

[0041] Calculate the position of the mobile robot at the current coordinate X. k The net force acting on the body:

[0042] The repulsive force F from the obstacle at the current position of the mobile robot is calculated based on the modified repulsive potential field sampling function. rep (X k ).

[0043] F rep (X k By considering the repulsive potential field function U rep (X k The negative gradient is obtained, and its physical meaning represents the repulsive potential field function U. rep (X k The steepest descent direction of the inspection robot requires the distance ρ(X) from the obstacle to the inspection robot. k The derivative of (x, x0) is given by the following mathematical expression:

[0044]

[0045]

[0046] Conclusion:

[0047]

[0048] Since the repulsive potential field function is a sampling space in this paper, n sets of repulsive forces will be obtained, which can be represented by the set F. rep (X k )={f rep1 (X k ),f rep2 (X k )……f repn (X k )}express.

[0049] The mobile robot is subjected to an attractive potential field of U from the target. att (X k The method for calculating the attractive potential field is as follows:

[0050]

[0051] F att (X k By considering the repulsive potential field function U att (X k The negative gradient is obtained, and its physical meaning represents the repulsive potential field function U. att (X k The steepest descent direction of the inspection robot requires the distance ρ(X) from the inspection robot to the target point. k X g) and its mathematical expression and brief process are as follows:

[0052]

[0053] The resultant force F(X k ) of the mobile robot is calculated k ), which is the vector resultant of the repulsive force F rep (X k ) and the attractive force F att (X k ).

[0054] F(X k ) = F att (X k ) + F rep (X k )

[0055] Since the repulsive force suffered by the mobile robot is a set F rep (X k ) = {f rep1 (X k ), f rep2 (X k ) …… F repn (X k )}, the resultant force can be expressed as a set F(X k ) = {f1(X k ), f2(X k ) …… f n (X k )}

[0056] 2. Calculate the force of the subsequent point and get the trajectory set.

[0057] After calculating F(X k ), set the step value s, and move the mobile robot in the direction of the F(X k ) vector by s steps to get the position coordinates X k+1 (x, y) of the k+1 point. Calculate the subsequent point X k+1 (x, y), X k+2 (x, y) …… X k+m (x, y) in the same way. X k+m (x, y) is the end point coordinate of a single local prediction trajectory, and the resultant force F(X k+1 ), F(X k+2 ) …… F(X k+m ) of the subsequent point is calculated. In the subsequent calculation, the sampling process is not used, and q takes the value

[0058] Given the same repulsive potential field function, the line connecting adjacent points is the path TRA that the inspection robot will traverse, also represented by a set TRA = {TRA1, TRA2, ..., TRA}. n The number of trajectories is the same as the number of q-value samples.

[0059] 3. Construct the evaluation function.

[0060] After the robot's path trajectory is generated, the optimal path needs to be selected based on the evaluation function, which mainly consists of the following evaluation indicators.

[0061] (1) Azimuth evaluation function heading(F att (X k ),F rep (X k The vector direction of the trajectory endpoint coordinates is parallel to the current position coordinates X. k (x, y) and the position coordinates of the target point X g The angle Δθ between the lines connecting (x, y) is used as the evaluation index. In this paper, to ensure that a larger evaluation function indicates a better path, π - Δθ is chosen. Therefore, heading(F att (X k ),F rep (X k ))=π-Δθ.

[0062] (2) Steepest descent evaluation function velocity(F) att (X k ),F rep (X k Its evaluation index is the coordinate position of the trajectory endpoint. Dist(V) is the Euclidean distance to the endpoint of the trajectory generated by the resultant force acting on the current position of the mobile robot. Since this paper adopts the evaluation method that the higher the evaluation score, the better the path, the output evaluation index is marked as -Dist(V).

[0063] (3) Local Minimum Avoidance Evaluation Function minima(F) att (X k ),F rep (X k Its evaluation indicators are divided into the following two situations.

[0064] When F(X) appears on the path k When the node is 0, minimal(F) att (X k ),F rep (X k ))=-∞, when F(X) does not exist in the path. k) = 0, minima(F att (X k ), F rep (X k )) = 0.

[0065] The evaluation functions heading(F att (X k ), F rep (X k )) and velocity(F att (X k ), F rep (X k )) are normalized.

[0066]

[0067] Where i represents the i-th simulated trajectory.

[0068] The evaluation functions are mathematically expressed as follows.

[0069] E v (F att (X k ), F rep (X k )) = σ [αheading(F att (X k ), F rep (X k )) ] + σ [βvelocity(F att (X k ), F rep (X k )) ] + σ [γminima(F att (X k ), F rep (X k ))]

[0070] Wherein α, β, γ are coefficients of the evaluation functions.

[0071] The final evaluation scores E v (F att (X k ), F rep (X k )) of the n trajectories are compared, and the one with the highest score is the optimal path selected in one cycle.

[0072] The mobile robot moves along the selected optimal path with a step size s, and when it moves to the midpoint of the local path, the coordinates X k and X g of the current point and the target point are calculated.Euclidean distance dist(X k ,X g ) :

[0073]

[0074] If dist(X k ,X g )<s, it is determined that the target point is reached, the obstacle avoidance process ends, and if dist(X k ,X g )>s, the loop process of calculating the resultant force received by the current position coordinates is re-entered, and the movement towards the target point continues.

Claims

1. A method for obstacle avoidance of a patrol robot based on fusion of artificial potential field method and dynamic window method, characterized in that The method comprises the following steps: (1) Modify the repulsive potential field function: modify the repulsive potential field function to a sampling interval U rep (X k ) rep1 (X k ) rep2 (X k ) repn (X k ) rep (X k ) rep1 (X k ) rep2 (X k ) repn (X k ) , where U rep (X k ) is the repulsive potential field function of the inspection robot at X k , F rep (X k ) is the repulsive force of the inspection robot at X k ; (2) generate a plurality of prediction routes: according to the attraction force from the target and the repulsion force from the obstacles received by the mobile robot, the resultant force F(X k ) received is derived = {f1(X k ), f2(X k )……f n (X k )}, the resultant force of the subsequent nodes is calculated, at this time, the resultant force of the subsequent nodes does not need to be sampled, and the node position coordinate change trajectory is the path trajectory TRA = {TRA1, TRA2……TRA n |; (3) Selecting the optimal path by the evaluation function: define the evaluation function by the constraint condition, including the angle with the terminal position, the steepest descent direction and the stress condition of the next point to score, and the TRA with the highest score is the optimal path for the robot to take.

2. The obstacle avoidance method for a patrol robot based on fusion of artificial potential field method and dynamic window method according to claim 1, characterized in that: In step 1, the repulsive potential field function is modified into a sampling interval, and the specific method is as follows: where k rep represents the repulsion gain coefficient, X k represents the position coordinates of the mobile robot, X0 represents the position coordinates of the obstacle, p0 represents the maximum influence distance of the obstacle repulsion potential field, p(X k , X0) = ||X0-X|| is the Euclidean distance between the mobile robot and the obstacle, q represents a randomly selected variable in the interval (0, 1), and is represented by the set q = {q1, q2, q3, …}, n samples are drawn from q by random sampling to obtain n different repulsion potential field functions.

3. The obstacle avoidance method for inspection robots based on the fusion of artificial potential field method and dynamic window method as described in claim 1, wherein the repulsive force set F is obtained in step 1. rep (X k )={f rep1 (X k ), f rep2 (X k )……f repn (X k The specific method is as follows: repulsive force F rep (X k ) is obtained by taking the negative gradient of the repulsive potential field function U rep (X k ), which represents the steepest descent direction of the repulsive potential field function U rep (X k ), the derivative of the distance between the inspection robot and the obstacle ρ(X k , X0) needs to be taken, and its mathematical expression and brief process are as follows: It is concluded that: Since the repulsive potential field function is modified as a sample space satisfying the random sampling condition, the repulsive force received by the inspection robot can be expressed as a set F rep (X k ) = {f rep1 (X k ), f rep2 (X k ), …, f repn (X k )}.

4. The obstacle avoidance method for a patrol robot based on fusion of artificial potential field method and dynamic window method according to claim 1, characterized in that: The resultant force set F(X k ) received by the mobile robot in step 1 k ) = {f1(X k ), f2(X n ) … fn(X k )}, and the calculation method is as follows: The attractive potential field from the target for the mobile robot is U att (X k The calculation method of the attractive potential field is as follows F att (X k ) is obtained by taking the negative gradient of the repulsive potential field function U att (X k ), which has the physical meaning of the steepest descent direction of the repulsive potential field function U att (X k ). The distance p(X k , X g ) between the inspection robot and the target point needs to be derived, and its mathematical expression and brief process are as follows: k att is a positive proportional gain factor of the attractive potential field, X g is the position coordinate of the target point, and the resultant force F(X k ) that the mobile robot receives is calculated as follows: k F(X F(X k ) = F att (X k ) + F rep (X k ) Since the repulsive forces experienced by the mobile robot are a set F rep (X k ) = {f rep1 (X k ), f rep2 (X k ),... f repn (X k )}, the resultant force experienced can be represented as a set F(X k ) = {f1(X k ), f2(X k ),... f n (X k )}.

5. The obstacle avoidance method for a patrol robot based on the fusion of artificial potential field method and dynamic window method according to claim 1, characterized in that: In step 2, a plurality of predicted routes are generated, and the specific method is as follows: According to the attraction force F from the target received by the mobile robot att (X k ) and the elements f in the repulsion force set F rep (X k ) rep1 (X k ) of the mobile robot X k point received the resultant force F(X k ), the subsequent point X k+1 , X k+2 ……X k+m received the resultant force F(X k+1 ), F(X k+2 )……F(X k+m ), at this time the calculation of the resultant force does not need to be sampled, that is, the resultant force is unique, q value Set the step value s, the line between adjacent point positions under the same repulsion potential field function premise is the path trajectory TRA that the inspection robot will take, also represented by a set TRA={TRA1, TRA2……TRA n}.

6. The obstacle avoidance method for a patrol robot based on fusion of artificial potential field method and dynamic window method according to claim 1, characterized in that: In step 3, the optimal path is selected by the evaluation function, and the specific method is as follows: After generating the trajectory, the multiple groups of trajectories obtained by sampling need to be evaluated and selected, each path is scored and evaluated by the evaluation standard, and the path trajectory with the highest score is selected as the actual motion trajectory of the inspection robot. The mathematical expression of the evaluation function is as follows: E v (F att (X k ), F rep (X k )) = σ [αheading(F att (X k ), F rep (X k )) ] + σ [βvelocity(F att (X k ), F rep (X k )) ] + σ [γminima(F att (X k ), F rep (X k ))] Wherein, α, β, γ are all coefficients of the evaluation function, since the process of local path planning needs to collect information by multiple sensors, the collected information cannot be continuous, which will also make the difference after evaluation larger, so normalization processing (smoothing processing) can be carried out, wherein σ represents normalization.

7. The obstacle avoidance method for a patrol robot based on the fusion of artificial potential field method and dynamic window method according to claim 6, characterized in that: The normalization processing process is as follows: Wherein, i represents the i th simulated trajectory.

8. The obstacle avoidance method for a patrol robot based on fusion of artificial potential field method and dynamic window method according to claim 5, characterized in that The score item of the evaluation function is: (1) heading(F att (X k ), F rep (X k )) is the azimuth angle evaluation function, the purpose is to avoid the generation of redundant nodes in the path of the inspection robot due to the repulsion force of the obstacles, and the included angle Δθ of the line connecting the direction of the trajectory end position and the target point is taken as the evaluation index. In order to make the evaluation function larger, indicating that the path is better, π-Δθ is selected, so heading(F att (X k ), F rep (X k )) = π-Δθ; (2) velocity (F att (X k ), F rep (X k )) is the steepest descent evaluation function, which aims to adjust the path deviation of the inspection robot caused by the influence of other evaluation functions in the evaluation, avoid deviating from the steepest descent direction, and the evaluation index is the Euclidean distance Dist(V) between the coordinate position of the trajectory endpoint and the coordinate position of the trajectory endpoint at the time, since this paper adopts the evaluation method that the higher the evaluation score, the better the path, so the output evaluation index is marked as -Dist(V); (3) minima(F att (X k ), F rep (X k )) is a local minimum value avoidance evaluation function, when the resultant force of the node on the path is 0, the path is discarded, so as to avoid falling into the local minimum value point through the evaluation function before falling into the local minimum value, and the evaluation index is F(X k ), in order to meet the evaluation mode that the higher the total score is, the better the path is, when F(X k ) is 0, minima(F att (X k ), F rep (X k )) = -∞, when there is no point with F(X k ) of 0 on the path node, minima(F att (X k ), F rep (X k )) = 0; Finally, the path with the highest score is selected. When the mobile robot reaches the end of the local path, it re-enters the sampling process and enters the next local path planning procedure. This step is repeated until a point X k The Euclidean distance between the target and the point X is less than the step size s, i.e., it is determined that the target point of the local path planning is reached, and the obstacle avoidance process is completed.

Images