Sweeping robot area coverage path planning method based on improved fishing optimization algorithm

Abstract

The invention discloses a sweeping robot area coverage path planning method based on an improved fishing optimization algorithm, and the method comprises the steps: obtaining environment information through a laser radar and / or a camera, and constructing a grid map; improving a fishing optimization algorithm; and inputting the created path grid map into the improved fishing optimization algorithm, and outputting an optimal coverage path. The method can be more effectively applied to area coverage path planning of the sweeping robot, and the coverage efficiency and safety of the robot in a complex environment are improved.

Description

Technical Field

[0001] The present invention relates to a method for path planning of area coverage of a floor cleaning robot based on an improved fishing optimization algorithm. Background Art

[0002] Floor cleaning robots complete floor cleaning in an automated manner, which can greatly save people's time and energy. However, to enable floor cleaning robots to work efficiently in complex indoor environments and achieve full coverage cleaning, the path planning problem has become one of its core technologies. The main problem that floor cleaning robots need to solve during path planning is how to quickly and effectively find a path that covers the entire cleaning area according to the complexity of the environment. Due to the dynamics and uncertainties of the indoor environment, such as furniture, obstacles, etc., traditional path planning algorithms often have difficulty dealing with these complex factors. Therefore, how to design an intelligent, adaptable and efficient path planning algorithm has become the key to improving the working efficiency of floor cleaning robots.

[0003] Currently, the research on path planning of floor cleaning robots mainly focuses on two categories: traditional algorithms and intelligent optimization algorithms. Traditional path planning methods such as the A* algorithm, Dijkstra algorithm, etc. usually rely on known environmental information and determine the shortest path or optimal path through graph search. However, these algorithms often perform inflexibly when faced with dynamically changing environments and complex obstacles, lacking real-time performance and adaptability. To solve the limitations of traditional algorithms, in recent years, bio-inspired algorithms have gradually become a hot topic in path planning research. Bio-inspired algorithms can perform effective global search in complex environments by simulating the behaviors of biological populations in nature, and have strong adaptive capabilities and global optimization capabilities. Common bio-inspired algorithms such as the particle swarm algorithm, ant colony algorithm, and genetic algorithm, etc. These algorithms optimize the search process by simulating the interactions between individuals, thereby finding the global optimal solution. In the path planning of floor cleaning robots, the application of these algorithms can effectively avoid the trap of local optimal solutions and improve the global search ability and efficiency of path planning. Summary of the Invention

[0004] The present invention provides a method for path planning of area coverage of a floor cleaning robot based on an improved fishing optimization algorithm to solve the problems existing in the above-mentioned prior art.

[0005] The technical solutions adopted by the present invention are as follows:

[0006] A method for path planning of area coverage of a floor cleaning robot based on an improved fishing optimization algorithm, comprising the following steps:

[0007] S1: Use a lidar and / or a camera to obtain environmental information and construct a grid map;

[0008] S2: Improve the fishing optimization algorithm, and the improvement points are as follows:

[0009] 1. In the initialization stage, initialize the positions of each fisherman by introducing Latin hypercube sampling;

[0010] 2. Introduce a dynamic capture rate mechanism based on search state feedback to adjust the capture rate;

[0011] 3. In the independent search stage, use Lévy flight to increase the search diversity of fishermen;

[0012] 4. In the group capture and collective capture, retain the optimal solution in each generation of fishermen through the designed elite-dominated strategy, and use the retained optimal solution to guide the search direction in the subsequent search process;

[0013] S3: Input the path raster map created in S1 into the improved fishing optimization algorithm, and output the optimal coverage path.

[0014] Furthermore, initialize the positions of each fisherman in each dimension by introducing Latin hypercube sampling, so that the initial fishermen are evenly distributed in the search space. The formula is defined as:

[0015] x i,j =lb j +(ub j -lb j )·LHS(dim,i)

[0016] In the formula, x i,j represents the initial position of the i-th fisherman in the j-th dimension; lb j is the lower limit of the search range; ub j is the upper limit of the search range; LHS(dim,i) represents the Latin hypercube sampling value of the i-th fisherman in the j-th dimension.

[0017] Furthermore, the dynamic capture rate mechanism based on search state feedback is specifically as follows: In the exploration stage of the fishing optimization algorithm, when p < α dynamic , fishermen tend to search independently. When p ≥ α dynamic , fishermen tend to group capture and collective enclosure; p is a random number between (0,1), and the specific formula for the dynamic capture rate α dynamic is as follows:

[0018]

[0019] In the formula, α initial is the initial capture rate; EFs is the current iteration number; MaxEFs is the maximum iteration number; Δfit is the change in the optimal fitness value in the current iteration; fit max and fit minThey are the maximum fitness value and the minimum fitness value in the current iteration respectively; is the best fitness value after T iterations; is the best fitness value after T-1 iterations.

[0020] Furthermore, in the independent search stage, the exploration ability of each fisherman is enhanced by adding Levy flight. After adding Levy flight, the position update formula of each fisherman is as follows:

[0021]

[0022] In the formula, is the position of the i-th fisherman at the (T + 1)-th iteration in the j-th dimension; is the position of the i-th fisherman at the T-th iteration in the j-th dimension; is the optimal fisherman position at the (T + 1)-th iteration in the j-th dimension; Exp is the empirical analysis value obtained by the fisherman as a reference object, and the range of this value is (-1, 1); rs is a random number between [0, 1]; s is a random unit vector with dimension d; d is the dimension; R is the movement radius of the fisherman; Levy(β) is the Levy flight step size, and β is the parameter controlling the step size.

[0023] Furthermore, the formula of the elite-dominated strategy is:

[0024]

[0025] In the formula, c is a group of 3 - 4 fisherman individuals whose positions are not updated; Centre c is the target point surrounded by group c;

[0026] is the position of the i-th fisherman in group c at the (T + 1)-th iteration in the j-th dimension; is the position of the i-th fisherman in group c at the T-th iteration in the j-th dimension; r 2 is the speed at which the fisherman approaches the target point surrounded by group c, and the value range is (0, 1); r 3 is the movement offset, and the value range is (-1, 1); α is the parameter controlling the influence of the elite solution; Elite best is the current elite solution; Fisher i T+1 is the position of the i-th fisherman at the (T + 1)-th iteration; Gbest is the global optimal position; GD is the Gaussian distribution function, the overall mean is 0, and σ is the overall variance; mean(Fisher) is the matrix of the average values of each dimension of the fisherman centered on the global optimal position; r 4 is a random number in {1, 2, 3}, distributed in three ranges; Fisher i T is the position of the i-th fisherman at the T-th iteration.

[0027] The present invention has the following beneficial effects:

[0028] In the population initialization stage of the Improved Catching Fish Optimization Algorithm (ICFOA), Latin Hypercube Sampling (LHS) is used to initialize the positions of fishermen, ensuring the uniformity and diversity of the initial population in the search space, avoiding the population aggregation problem that may be caused by traditional random initialization, enhancing the global search ability of the algorithm, and enabling it to find potential high-quality solutions more quickly in complex environments. In the exploration stage, a dynamic capture rate mechanism based on search state feedback is introduced, allowing the algorithm to flexibly adjust strategies according to the current search state, improving the efficiency and flexibility of the search. In the independent search stage, the Lévy flight strategy is adopted, enhancing the search diversity through the combination of long and short jumps, avoiding the algorithm falling into local optimal solutions, and thus more effectively exploring unknown areas. In the group catching and collective catching stages, an elite-dominated strategy is adopted, accelerating the convergence speed of the algorithm and improving the accuracy and stability of path planning. The present invention can be more effectively applied to the area coverage path planning of a sweeping robot, improving the coverage efficiency and safety of the robot in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of the present invention.

[0030] Figure 2 is a flow chart of the improved catching fish optimization algorithm.

[0031] Figure 3a is f 3 optimization function.

[0032] Figure 3b is a convergence curve.

[0033] Figure 4a is f 5 optimization function.

[0034] Figure 4b is a convergence curve.

[0035] Figure 5a is f 6 optimization function.

[0036] Figure 5b is a convergence curve.

[0037] Figure 6a is f 7 optimization function.

[0038] Figure 6b is a convergence curve.

[0039] Figure 7a is f 10 optimization function.

[0040] Figure 7b is the convergence curve.

[0041] Figure 8a is f 15 optimization function.

[0042] Figure 8b is the convergence curve. Detailed implementation manners

[0043] The present invention will be further described below with reference to the accompanying drawings.

[0044] Such as Figure 1 and Figure 2 , a method for path planning of area coverage of a floor cleaning robot based on an improved fishing optimization algorithm according to the present invention includes the following steps:

[0045] S1: Use a lidar and / or a camera to obtain environmental information and construct a grid map;

[0046] S2: Improve the fishing optimization algorithm, and the improvement points are:

[0047] (1) Introduce Latin hypercube sampling (LHS) to initialize the positions of fishermen, ensuring a more uniform initial distribution of fishermen in the search space. This method overcomes the problem of population aggregation that may be caused by traditional random initialization, enables uniform coverage of all regions in the search space, and avoids insufficient exploration of some regions. This improvement effectively enhances the global search ability of the algorithm, improves the efficiency of finding the global optimal solution in complex problems, and enables the algorithm to better explore efficiently in a vast search space.

[0048] (2) Introduce a dynamic capture rate mechanism based on search state feedback in the fishing optimization algorithm. The algorithm can intelligently adjust the capture rate according to the current search state. When p < α dynamic , the capture rate is high, promoting fishermen to quickly explore unknown regions. When p ≥ α dynamic , the capture rate is low, enabling fishermen to pay more attention to local regions, thereby strengthening fine-grained search.

[0049] (3) Adopt Lévy flight in the independent search stage to increase the diversity of fishermen's search. It effectively avoids the algorithm from falling into local optimal solutions and enables the algorithm to better explore unknown regions in global search, thereby improving the comprehensiveness and accuracy of search.

[0050] (4) In the group capture and exploitation stage, retain the optimal solutions among fishermen in each generation through the designed elite-dominated strategy, and use these elite solutions to guide the search direction in subsequent search processes.

[0051] S3: Input the path raster map created in S1 into the improved fishing optimization algorithm to output the optimal coverage path.

[0052] Regarding improvement point 1:

[0053] Initialize the positions of each fisherman in each dimension through Latin hypercube sampling (LHS) so that the initial fishermen are evenly distributed in the search space, enhancing diversity. The formula is defined as:

[0054] x i,j = lb j +(ub j - lb j )·LHS(dim, i)

[0055] In the formula, x i,j represents the initial position of the i-th fisherman in the j-th dimension; lb j is the lower limit of the search range; ub j is the upper limit of the search range; LHS(dim, i) represents the Latin hypercube sampling value of the i-th fisherman in the j-th dimension.

[0056] Regarding improvement point 2:

[0057] In the exploration stage of the fishing optimization algorithm, when p < α dynamic , the capture rate is relatively high, promoting fishermen to quickly explore unknown areas; when p ≥ α dynamic , the capture rate is relatively low, enabling fishermen to pay more attention to local areas, thus strengthening the refined search; this mechanism enables the algorithm to flexibly adjust strategies at different stages, avoid over-exploring irrelevant areas, improve the search efficiency and the quality of the final solution. The specific formula for the dynamic capture rate α dynamic is as follows:

[0058]

[0059] In the formula, α initial is the initial capture rate; EFs is the current iteration number; MaxEFs is the maximum iteration number; Δfit is the change in the optimal fitness value in the current iteration; fit max and fit min are the maximum and minimum fitness values in the current iteration respectively; is the best fitness value after T iterations; is the best fitness value after T - 1 iterations.

[0060] Regarding improvement point 3:

[0061] In the independent search stage, fishermen rely on local search and random perturbation to explore the waters, and the search efficiency is relatively low. The exploration ability is enhanced by adding Levy flight. Levy flight enhances the search diversity of fishermen through the combination of long jumps and short jumps. Compared with the traditional simple random search method, Levy flight can conduct fine search in the local area while also jumping to a completely new area for large-scale exploration. After adding Levy flight, the position update formula for each fisherman is as follows:

[0062]

[0063] In the formula, is the position of the i-th fisherman at the (T + 1)-th iteration in the j-th dimension; is the position of the i-th fisherman at the T-th iteration in the j-th dimension; is the optimal fisherman position at the (T + 1)-th iteration in the j-th dimension; Exp is the empirical analysis value obtained by the fisherman as a reference object, and the range of this value is (-1, 1); rs is a random number between [0, 1]; s is a random unit vector with the dimension of d; d is the dimension; R is the movement radius of the fisherman; Levy(β) is the Levy flight step size, and β is the parameter controlling the step size.

[0064] Regarding improvement point 4:

[0065] During group capture and collective capture, the cooperation and collective capture strategies of fishermen will lead to a limited search range. When approaching the optimal solution, fishermen concentrate in one area. The elite-dominated strategy ensures that the optimal solution will not be lost during the search process by retaining the optimal fishermen in each generation. The elite solution not only serves as the representative of the current optimal solution but also plays a guiding role in subsequent searches, helping fishermen avoid deviating from the optimal path in terms of direction, improving the accuracy and stability of path planning, avoiding possible search degradation, and enhancing the solution ability and reliability of the algorithm in complex environments. The position update formula is as follows:

[0066]

[0067] In the formula, c is a group of 3 - 4 fishermen individuals whose positions have not been updated; Centre c is the target point surrounded by group c; is the position of the i-th fisherman in group c at the (T + 1)-th iteration in the j-th dimension; is the position of the i-th fisherman in group c at the T-th iteration in the j-th dimension; r 2 is the speed at which the fisherman approaches the target point surrounded by group c, and the value range is (0, 1); r 3 is the offset of the movement, and the value range is (-1, 1); α is the parameter controlling the influence of the elite solution; Elite best is the current elite solution; Fisher i T+1is the position of the i-th fisherman at the (T+1)-th iteration; Gbest is the global optimal position; GD is the Gaussian distribution function with a population mean of 0 and σ being the population variance; mean(Fisher) is the matrix of the average values of each dimension of the fishermen centered around the global optimal position; r 4 is a random number in {1, 2, 3}, distributed within three ranges; Fisher i T is the position of the i-th fisherman at the T-th iteration.

[0068] Figure 3a is f 3 the optimization function, Figure 3b is the convergence curve. Figure 3a Displays the 3D landscape of the values, usually used to visualize multi-dimensional objective functions. Different colors and heights in the figure represent different levels of the objective function. Peaks and valleys represent the optimization regions that the algorithm attempts to minimize or maximize. Figure 3b Shows how the objective function values of four different optimization algorithms decrease during the iteration process. Among them, ICFOA performs the best, and the objective function value decreases the fastest.

[0069] Figure 4a is f 5 the optimization function, Figure 4b is the convergence curve. Figure 4a Has two different valleys or regions where the function values change drastically. The gradient of the color represents the range of the objective function values. Due to the steep gradient and large value range of this function, it is very challenging for optimization algorithms. It can be seen from Figure 4b that ICFOA is again superior to other algorithms in terms of convergence speed and stability, reaching a lower value in fewer iterations.

[0070] Figure 5a is f 6 the optimization function, Figure 5b is the convergence curve. Figure 5a Has a complex terrain with valleys and ridges. The color gradient represents different function values, with yellow areas representing higher values and blue areas representing lower values. This function has multiple local minima, which makes it difficult for optimization algorithms to effectively converge to the global minimum, but Figure 5b the ICFOA in

[0071] Figure 6a is f 7 the optimization function, Figure 6b is the convergence curve. Figure 6ais a highly non-convex function with multiple sharp peaks and valleys; yellow represents higher values ​​and blue represents lower values. The sharp changes in the surface indicate the presence of multiple local minima, which can trap the optimization algorithm in a suboptimal solution, but Figure 6b ICFOA in has always maintained its leading position in terms of convergence speed and stability, and is able to quickly reach lower objective function values ​​and maintain that performance.

[0072] Figure 7a f 10 Optimize functions, Figure 7b is the convergence curve. Figure 7a The surface plot shows a steep cone with a very sharp peak in the center. The function has a very narrow and deep valley and high sensitivity around the optimal point, making it a challenging optimization problem. The color gradient shows that the value of the function drops sharply as you move toward the center, and there is a strong local minimum. Figure 7b ICFOA in again shows the fastest convergence rate, with the objective function value decreasing dramatically and stabilizing at a very low value after about 200 iterations.

[0073] Figure 8a f 15 Optimize functions, Figure 8b is the convergence curve. Figure 8a The function has a two-peaked shape, so it has multiple local maxima. This function is highly non-convex, and the gradient near the peak is steep, which makes it difficult for the optimization algorithm to move towards the global optimum. Figure 8b It can be seen that ICFOA shows the fastest convergence speed, quickly reaches a very low objective function value and stabilizes at the minimum value.

[0074] In general, the three-dimensional graph on the left side of each figure shows the typical optimization function and its characteristics, which is designed to test the performance of the algorithm on different problems; the convergence curve on the right side shows the trend of the fitness value of the four algorithms with the number of iterations. It can be seen from the figure that the improved fishing optimization algorithm represented by the purple curve is better than other algorithms in terms of convergence speed and final fitness value, indicating that the improved version of the fishing optimization algorithm has better performance in solving optimization problems.

[0075] Table 1 gives the test results of each algorithm on different optimization functions.

[0076] Table 1 Function comparison experimental results

[0077]

[0078] According to Table 1, the bold values in Table 1 represent the optimal results, showing that the ICFOA algorithm is significantly superior to other comparison algorithms in terms of the optimal value, average value, and standard deviation. Especially in terms of accuracy, convergence speed, the ability to escape local optima, and overall stability, ICFOA demonstrates excellent performance. Specifically, the ICFOA algorithm can converge to the optimal solution more quickly and has a stronger global search ability in complex environments, avoiding being trapped in local optima and ensuring the accuracy and stability of the solution. These excellent performances indicate that the ICFOA algorithm has strong adaptability and robustness when facing complex scenarios. Generally speaking, the advantages of the ICFOA algorithm in terms of global search ability, convergence accuracy, and stability make it perform well in solving the complex area coverage path planning problem.

[0079] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be regarded as the protection scope of the present invention.

Claims

1. A method for area coverage path planning of a sweeping robot based on an improved fishing optimization algorithm, characterized in that: The steps include: S1: Use lidar and / or camera to obtain environmental information and build a grid map; S2: Improve the fishing optimization algorithm. The improvements are as follows:

1. In the initialization stage, the position of each fisherman is initialized by introducing Latin hypercube sampling; 2. Introduce a dynamic capture rate mechanism based on search status feedback to adjust the capture rate; 3. Use Levy flights to increase fisherman search diversity during the independent search phase; 4. During group capture and collective capture, the optimal solution of each generation of fishermen is retained through the designed elite-dominated strategy, and the retained optimal solution is used to guide the search direction in the subsequent search process; S3: Input the path grid map created by S1 into the improved fishing optimization algorithm and output the optimal coverage path.

2. The method for area coverage path planning of a sweeping robot based on an improved fishing optimization algorithm as claimed in claim 1, characterized in that: By introducing Latin hypercube sampling to initialize the position of each fisherman in each dimension, the initial fishermen are evenly distributed in the search space. The formula is defined as: x i,j =lb j +(ub j -lb j )·LHS(dim,i) In the formula, x i,j represents the initial position of the i-th fisherman in the j-th dimension; lb j is the lower limit of the search range; ub j is the upper limit of the search range; LHS(dim,i) represents the Latin hypercube sampling value of the i-th fisherman in the j-th dimension.

3. The method for area coverage path planning of a sweeping robot based on an improved fishing optimization algorithm as claimed in claim 1, characterized in that: The dynamic capture rate mechanism based on search state feedback is as follows: in the exploration phase of the fishing optimization algorithm, when p < α dynamic When p ≥ α, fishermen tend to search independently. dynamic When , fishermen tend to capture in groups and collectively; p is a random number between (0,1), and the dynamic capture rate α dynamic The specific formula is as follows: In the formula, α initial is the initial capture rate; EFs is the current number of iterations; MaxEFs is the maximum number of iterations; Δfit is the change in the optimal fitness value in the current iteration; fit max and fit min They are the maximum fitness value and the minimum fitness value in the current iteration respectively; It is the optimal fitness value after T iterations; is the optimal fitness value after iteration T-1 times.

4. The method for area coverage path planning of a sweeping robot based on an improved fishing optimization algorithm as claimed in claim 1, characterized in that: In the independent search phase, the exploration ability of each fisherman is enhanced by adding Levi flights. After adding Levi flights, the position update formula of each fisherman is as follows: In the formula, is the position of the i-th fisherman at the j-th dimension T+1 iteration; is the position of the i-th fisherman in the j-th dimension at the T-th iteration; is the optimal fisherman position in the jth dimension T+1 iteration; Exp is the empirical analysis value obtained with the fisherman as the reference object, and the value range is (-1,1); rs is a random number between [0, 1]; s is a random unit vector with dimension d; d is the dimension; R is the movement radius of the fisherman; Levy (β) is the Levy flight step length, and β is the parameter that controls the step length.

5. The method for area coverage path planning of a sweeping robot based on an improved fishing optimization algorithm as claimed in claim 1, characterized in that: The formula for the elite-dominated strategy is: Where c is a group of 3-4 fishermen whose positions have not been updated; Centre c is the target point surrounded by group c; is the position of the i-th fisherman in group c at the j-th dimension T+1 iteration; is the position of the i-th fisherman in group c at the j-th iteration of dimension T; r2 is the speed of the fisherman approaching the target point surrounded by group c, and its value is (0, 1); r3 is the offset of the movement, and its value is (-1, 1); α is the parameter that controls the influence of the elite solution; Elite best is the current elite solution; Fisher i T+1 is the position of the i-th fisherman at the T+1 iteration; Gbest is the global optimal position; GD is a Gaussian distribution function with a population mean of 0 and σ being the population variance; mean(Fisher) is the matrix of the average values ​​of each dimension of the fisherman centered at the global optimal position; r4 is a random number {1, 2, 3} distributed in three ranges; Fisher i T is the position of the ith fisherman at T iterations.

Images