├── 1-Non-Obstacles ├── 10-10pathLength.xlsx ├── 20-17pathLength.xlsx ├── ACOPathPlanning_V1.m ├── ConvertXY.m ├── Note.md ├── calcHeuristic.m ├── delta_r2s.m ├── eta.xlsx ├── globalDelta.xlsx ├── globalDelta_Bug.xlsx ├── indexing.m └── showPath.m ├── 2-LineObstacles ├── ACOPathPlanning.m ├── ConvertXY.m ├── Iterations.jpg ├── Log.md ├── Results │ ├── Length-Iteration__relativeImportance4.jpg │ ├── Length-Iteration_newPheromoneupdate.jpg │ ├── robotPath_newPheromoneupdate.jpg │ └── robotPath_relativeImportance4.jpg ├── calcDelta.m ├── calcHeuristic.m ├── indexing.m ├── robotPathRecords.jpg ├── showLength.m └── showPath.m ├── 3 - Convergence └── Log.md ├── 4 - Final Stage └── Log.md ├── README.md └── image ├── ACO Flowchart.png ├── ACOProjectLogo.png ├── Demo.gif ├── Map-Algorithm Boxchart.jpg ├── Map3.jpg ├── Map_Easy.png ├── Map_Hard.png ├── Min Path Length-Data (1).jpg ├── Min path length - data.jpg ├── ParaSelec_ACS_Mark.jpg ├── ParaSelec_MMAS_1.jpg ├── ParaSelec_eAS_1.png ├── PathPlanningAlgorithmsOverview.png ├── PerformanceAnalysis_CS.png ├── PerformanceAnalysis_Runtime.png ├── Route Map (1).jpg ├── Route Map (2).jpg ├── Route Map.jpg ├── aco_diagram.jpg ├── converge_map1.png ├── converge_map2.png ├── final_demo.gif ├── image_log.md ├── matlab.png ├── route_map1.png └── route_map2.png /1-Non-Obstacles/10-10pathLength.xlsx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/1-Non-Obstacles/10-10pathLength.xlsx -------------------------------------------------------------------------------- /1-Non-Obstacles/20-17pathLength.xlsx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/1-Non-Obstacles/20-17pathLength.xlsx -------------------------------------------------------------------------------- /1-Non-Obstacles/ACOPathPlanning_V1.m: -------------------------------------------------------------------------------- 1 | % Title: Ant Colony Optimisation algorithms for Robot Path Planning 2 | % Author: Yujin Li 3 | % Date: Sunday 13 March 2022 4 | clc; 5 | clear all; 6 | close all; 7 | % 1. INITIALISATION 8 | % 1.1. Environment 9 | SIZE = 10; 10 | gridMap = zeros(SIZE); % Environment configuration: 0 for clear path, 1 for obstacles 11 | % 1.2. Parameters: 12 | % pheromone, delta pheromone: inverse of the total path length, 13 | % heuristic values, number of ants, generations of ants, Alpha: evaporation 14 | % rate, Beta: relative importance of pheromone versus distance 15 | numAnts = 17; % the number of ants in a colony 16 | numGen = 20; % the number of generations (iterations) 17 | tau = ones(SIZE*SIZE); % Pheromone Matrix Tau(r,s): amount of pheromone on 18 | % the edge from grid r to grid s. In this case, the 19 | % first column represents all the grid r and the 20 | % first row represents all the grid s, where s must 21 | % belong to J_k(r), which is a set of cities that 22 | % remain to be visited by ant k positioned on city r 23 | 24 | tau = 8.*tau; % Set the initial pheromone of all edges (r,s) as 8 25 | alpha = 0.3; % the Evaporation rate of pheromone 26 | beta = 7; % the relative importance beta 27 | home = 1; % the starting point -- home of the robot 28 | destination = SIZE*SIZE; % destination location 29 | minTourLength = inf; % the length of the shortest path 30 | bestGen = 0; % the generation that finds the shortest path 31 | bestAnt = 0; % the ant that finds the shortest path within current generation 32 | globalDelta = delta_r2s(gridMap); % cost required from r to s 33 | % T_delta = table(globalDelta); 34 | % writetable(T_delta, 'globalDelta.xlsx', 'sheet', 1, 'Range', 'A1'); 35 | eta = zeros(SIZE); % Heuristic value Matrix Eta(r,s), s belongs to J_k(r) 36 | eta = calcHeuristic(eta, destination); % Initialise Heuristic value 37 | % T_eta = table(eta); 38 | % writetable(T_eta, 'eta.xlsx', 'Sheet', 1, 'Range', 'A1'); 39 | % edge = 1; % the size of the square grid( the length of the edge) 40 | pathStorage = cell(numGen, numAnts); 41 | pathLength = zeros(numGen, numAnts); 42 | 43 | % 2. STATE TRANSITION 44 | % 2.1 Initialisation 45 | for Gen = 1:numGen % Outter Loop: How many times ants go out and forage 46 | for Ant = 1:numAnts % Inner Loop: How many ants each time 47 | currGrid = home; % Store the current grid where the ant is located 48 | pathRecord = home; % Store the grids that the ant has walked through 49 | toBeVisited = ones(SIZE); % cities that remain to be visited by ant 50 | % k positioned on city r; 0 for visited, 51 | % 1 for to be visited 52 | toBeVisited(currGrid) = 0; % Mark the home grid as visited so it won't repeat 53 | localDelta = globalDelta; 54 | indexDelta = indexing(localDelta, currGrid, toBeVisited); % find indices of local available surrounding grids 55 | numAvailable = length(indexDelta); % the total number of the local available grids < 8 56 | counter = 1; 57 | % 2.2 The ant start finding path to destination 58 | while (currGrid ~= destination && numAvailable >= 1) 59 | % 2.2.1 Calculate the probability of choosing the certain grids based on 60 | % tau and eta 61 | stateTransProb = zeros(numAvailable, 1); 62 | for i = 1: numAvailable 63 | stateTransProb(i) = tau(currGrid, indexDelta(i)) * eta(indexDelta(i))^beta; 64 | end 65 | sumProb = sum(stateTransProb); 66 | stateTransProb = stateTransProb / sumProb; 67 | % 2.2.2 Roulette wheel selection algorithm: to choose the next grid randomly 68 | wheelProb = zeros(numAvailable, 1); 69 | wheelProb(1) = stateTransProb(1); 70 | for i = 2:numAvailable 71 | if i ~= numAvailable 72 | wheelProb(i) = wheelProb(i-1) + stateTransProb(i); 73 | else 74 | wheelProb(numAvailable) = 1; 75 | end 76 | end 77 | wheelHand = find(wheelProb >= rand, 1); % Turn the wheel and wait for the result 78 | nextGrid = indexDelta(wheelHand(1)); % Decide the next grid 79 | % 2.2.3 Store the result: the new grid, the updated length 80 | pathRecord(end + 1) = nextGrid; % Concatenate the new grid 81 | pathLength(Gen, Ant) = pathLength(Gen, Ant) + localDelta(currGrid, nextGrid); 82 | currGrid = nextGrid; % Move to the next grid 83 | % 2.2.4 Update J_k(r) and corresponding localDelta to prepare for the next move 84 | toBeVisited(currGrid) = 0; 85 | for i = 1:size(localDelta, 1) 86 | if toBeVisited(i) == 0 87 | localDelta(currGrid, i) = 0; 88 | localDelta(i, currGrid) = 0; 89 | end 90 | end 91 | indexDelta = indexing(localDelta, currGrid, toBeVisited); 92 | numAvailable = length(indexDelta); 93 | counter = counter + 1; 94 | end % the ant reaches to the destination and stop the loop 95 | % 2.3 Record the job done by the ant 96 | pathStorage{Gen, Ant} = pathRecord; % Save the path 97 | if pathRecord(end) == destination 98 | if pathLength(Gen, Ant) < minTourLength % Does the ant find a shorter path? 99 | minTourLength = pathLength(Gen, Ant); % Yes, save the new shortest path 100 | bestGen = Gen; % Remember the generation that achieved this 101 | bestAnt = Ant; % Remember the ant that achieved this 102 | end 103 | else % When the ant is not successful, offset its length in this iteration 104 | pathLength(Gen, Ant) = 0; 105 | end 106 | end % The ant finishes the forage this time 107 | % 2.4 Update the intensity of pheromone on the path 108 | deltaTau = zeros(size(tau)); 109 | for i = 1: numAnts 110 | if pathLength(Gen, Ant) 111 | pathTemp = pathStorage{Gen, Ant}; 112 | pathInterval = length(pathTemp) - 1; 113 | for j = 1: pathInterval 114 | deltaTau(pathTemp(j), pathTemp(j + 1)) = 1 / pathLength(Gen, Ant); 115 | deltaTau(pathTemp(j + 1), pathTemp(j)) = 1 / pathLength(Gen, Ant); 116 | end 117 | end 118 | end 119 | tau = (1 - alpha) .* tau + deltaTau; % pheromone evaporation rule 120 | end % All the ants in this generation finished forage; restart a new generation now 121 | 122 | % 3. VISUALISATION 123 | showPath(gridMap,SIZE, pathStorage, bestGen, bestAnt); 124 | -------------------------------------------------------------------------------- /1-Non-Obstacles/ConvertXY.m: -------------------------------------------------------------------------------- 1 | function [x, y] = ConvertXY(gridNum, rowMap) % Extract x coordinate from the serial number of the location 2 | x = ceil(gridNum / rowMap) - 0.5; 3 | y = rowMap - mod(gridNum, rowMap) + 0.5; 4 | if y == rowMap + 0.5 5 | y = 0.5; 6 | end 7 | end 8 | 9 | -------------------------------------------------------------------------------- /1-Non-Obstacles/Note.md: -------------------------------------------------------------------------------- 1 | This is the 1st working code 2 | -------------------------------------------------------------------------------- /1-Non-Obstacles/calcHeuristic.m: -------------------------------------------------------------------------------- 1 | function Eta = calcHeuristic(arr, grid) 2 | dim1 = size(arr, 1); 3 | dim2 = size(arr, 1) * size(arr, 1); 4 | for i = 1:dim2 5 | [desti_x, desti_y] = ConvertXY(grid, dim1); 6 | [x, y] = ConvertXY(i, dim1); 7 | distance = sqrt((x - desti_x)^2 + (y - desti_y)^2); 8 | if i ~= grid 9 | arr(i) = 1 / distance; 10 | else 11 | arr(grid) = 100; 12 | end 13 | end 14 | Eta = arr; 15 | end 16 | 17 | -------------------------------------------------------------------------------- /1-Non-Obstacles/delta_r2s.m: -------------------------------------------------------------------------------- 1 | function D = delta_r2s(arr) % delta_(r,s): cost from grid r to grid s 2 | % r and s must be next to each other ssss 3 | row = size(arr, 1); % the number of rows of the array 4 | column = size(arr, 2); 5 | D = zeros(row*column); % initialise the delta_(r,s) matrix 6 | % Outer Loop 7 | for i = 1:row 8 | for j = 1:column % Outter Loop: Traverse all the grids to set their cost 9 | % of traveling to the local surrouding grids 10 | if arr(i, j) == 0 % grid r must not be a obstacle 11 | 12 | % Inner Loop 13 | for x = 1:row 14 | for y = 1:column % Inner Loop: find the surrounding available 15 | % grids s and return the cost value 16 | % (distance) to the exact variable in 17 | % the matrix 18 | if arr(x, y) == 0 % the surrounding grid must be clear 19 | rel_x = abs(i - x); % relative distance between grid r and s on x direction 20 | rel_y = abs(j - y); % relative distance between grid r and s on y direction 21 | if rel_x + rel_y == 1 || (rel_x == 1 && rel_y == 1) 22 | % When this grid is one of the adjacent 23 | % grids in 8 directions 24 | D((i-1)*column + j, (x-1)*column + y) = sqrt(rel_x^2 + rel_y^2); 25 | % D(r, s), r and s are indices of one-dimensional matrix 26 | % Return relative distance between r and s 27 | end 28 | end 29 | end 30 | end 31 | end 32 | end 33 | end 34 | 35 | end 36 | 37 | -------------------------------------------------------------------------------- /1-Non-Obstacles/eta.xlsx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/1-Non-Obstacles/eta.xlsx -------------------------------------------------------------------------------- /1-Non-Obstacles/globalDelta.xlsx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/1-Non-Obstacles/globalDelta.xlsx -------------------------------------------------------------------------------- /1-Non-Obstacles/globalDelta_Bug.xlsx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/1-Non-Obstacles/globalDelta_Bug.xlsx -------------------------------------------------------------------------------- /1-Non-Obstacles/indexing.m: -------------------------------------------------------------------------------- 1 | function index = indexing(delta, grid, J_kr) % Return the indices of available grids 2 | index = delta(grid, :); % Store the cost required to the next gird 3 | arr = find(index); % Scope down available adjacent grids indices; store the complete indices 4 | % Double check for visited grids to avoid self-repeating 5 | for i = 1: length(arr) 6 | if J_kr(arr(i)) == 0 7 | index(arr(i)) = 0; 8 | end 9 | % What about excluding obstacles? grids == 1? 10 | % Why is the path going straight down? while heuristic value are 11 | end 12 | index = find(index); 13 | end 14 | 15 | -------------------------------------------------------------------------------- /1-Non-Obstacles/showPath.m: -------------------------------------------------------------------------------- 1 | function showPath(map, size, path, gen, ant) 2 | figure(1); 3 | for i = 1: size*size 4 | if map(i) == 1 5 | [x_p, y_p] = ConvertXY(i, size); 6 | x1 = x_p - 0.5; y1 = y_p + 0.5; 7 | x4 = x_p - 0.5; y4 = y_p - 0.5; 8 | x2 = x_p + 0.5; y2 = y_p + 0.5; 9 | x3 = x_p + 0.5; y3 = y_p - 0.5; 10 | fill([x1, x2, x3, x4], [y1, y2, y3, y4], [0.2 0.2 0.2]); 11 | hold on; 12 | else 13 | [x_p, y_p] = ConvertXY(i, size); 14 | x1 = x_p - 0.5; y1 = y_p + 0.5; 15 | x4 = x_p - 0.5; y4 = y_p - 0.5; 16 | x2 = x_p + 0.5; y2 = y_p + 0.5; 17 | x3 = x_p + 0.5; y3 = y_p - 0.5; 18 | fill([x1, x2, x3, x4], [y1, y2, y3, y4], 'w'); 19 | hold on; 20 | end 21 | end 22 | hold on; 23 | title('Map'); 24 | xlabel('x-axis'); 25 | ylabel('y-axis'); 26 | pathRecord = path{gen, ant}; 27 | ctr = length(pathRecord); 28 | robot_x = zeros(ctr, 1); 29 | robot_y = zeros(ctr, 1); 30 | for j = 1: ctr 31 | [robot_x(j), robot_y(j)] = ConvertXY(pathRecord(j), size); 32 | end 33 | plot(robot_x, robot_y); 34 | end 35 | 36 | -------------------------------------------------------------------------------- /2-LineObstacles/ACOPathPlanning.m: -------------------------------------------------------------------------------- 1 | % Title: Ant Colony Optimisation algorithms for Robot Path Planning 2 | % Author: Yujin Li 3 | % Date: Sunday 13 March 2022 4 | clc; 5 | clear all; 6 | close all; 7 | % 1. INITIALISATION 8 | % 1.1. Environment 9 | SIZE = 10; 10 | gridMap = zeros(SIZE); % Environment configuration: 0 for clear path, 1 for obstacles 11 | gridMap(56) = 1; 12 | % gridMap(9, 4:8) = 1; 13 | % gridMap(2:9, 5) = 1; 14 | % 1.2. Parameters: 15 | % pheromone, delta pheromone: inverse of the total path length, 16 | % heuristic values, number of ants, generations of ants, Alpha: evaporation 17 | % rate, Beta: relative importance of pheromone versus distance 18 | numAnts = 30; % the number of ants in a colony 19 | numGen = 50; % the number of generations (iterations) 20 | tau = ones(SIZE*SIZE); % Pheromone Matrix Tau(r,s): amount of pheromone on 21 | % the edge from grid r to grid s. In this case, the 22 | % first column represents all the grid r and the 23 | % first row represents all the grid s, where s must 24 | % belong to J_k(r), which is a set of cities that 25 | % remain to be visited by ant k positioned on city r 26 | 27 | tau = 8.*tau; % Set the initial pheromone of all edges (r,s) as 8 28 | alpha = 0.3; % the Evaporation rate of pheromone 29 | beta = 7; % the relative importance beta 30 | home = 1; % the starting point -- home of the robot 31 | destination = SIZE*SIZE; % destination location 32 | minTourLength = inf; % the length of the shortest path 33 | bestGen = 0; % the generation that finds the shortest path 34 | bestAnt = 0; % the ant that finds the shortest path within current generation 35 | globalDelta = calcDelta(gridMap); % cost required from r to s 36 | % T_delta = table(globalDelta); 37 | % writetable(T_delta, 'globalDelta.xlsx', 'sheet', 1, 'Range', 'A1'); 38 | eta = zeros(SIZE); % Heuristic value Matrix Eta(r,s), s belongs to J_k(r) 39 | eta = calcHeuristic(eta, destination); % Initialise Heuristic value 40 | % T_eta = table(eta); 41 | % writetable(T_eta, 'eta.xlsx', 'Sheet', 1, 'Range', 'A1'); 42 | % edge = 1; % the size of the square grid( the length of the edge) 43 | pathStorage = cell(numGen, numAnts); 44 | pathLength = zeros(numGen, numAnts); 45 | 46 | % 2. STATE TRANSITION 47 | % 2.1 Initialisation 48 | for Gen = 1:numGen % Outter Loop: How many times ants go out and forage 49 | for Ant = 1:numAnts % Inner Loop: How many ants each time 50 | currGrid = home; % Store the current grid where the ant is located 51 | pathRecord = home; % Store the grids that the ant has walked through 52 | toBeVisited = ones(SIZE); % cities that remain to be visited by ant 53 | % k positioned on city r; 0 for visited, 54 | % 1 for to be visited 55 | toBeVisited(currGrid) = 0; % Mark the home grid as visited so it won't repeat 56 | localDelta = globalDelta; 57 | indexDelta = indexing(localDelta, currGrid, toBeVisited); % find indices of local available surrounding grids 58 | numAvailable = length(indexDelta); % the total number of the local available grids < 8 59 | counter = 1; 60 | % 2.2 The ant start finding path to destination 61 | while (currGrid ~= destination && numAvailable >= 1) 62 | % 2.2.1 Calculate the probability of choosing the certain grids based on 63 | % tau and eta 64 | stateTransProb = zeros(numAvailable, 1); 65 | for i = 1: numAvailable 66 | stateTransProb(i) = tau(currGrid, indexDelta(i)) * eta(indexDelta(i))^beta; 67 | end 68 | sumProb = sum(stateTransProb); 69 | stateTransProb = stateTransProb / sumProb; 70 | % 2.2.2 Roulette wheel selection algorithm: to choose the next grid randomly 71 | wheelProb = zeros(numAvailable, 1); 72 | wheelProb(1) = stateTransProb(1); 73 | for i = 2:numAvailable 74 | if i ~= numAvailable 75 | wheelProb(i) = wheelProb(i-1) + stateTransProb(i); 76 | else 77 | wheelProb(numAvailable) = 1; 78 | end 79 | end 80 | wheelHand = find(wheelProb >= rand, 1); % Turn the wheel and wait for the result 81 | nextGrid = indexDelta(wheelHand(1)); % Decide the next grid 82 | % 2.2.3 Store the result: the new grid, the updated length 83 | pathRecord(end + 1) = nextGrid; % Concatenate the new grid 84 | pathLength(Gen, Ant) = pathLength(Gen, Ant) + localDelta(currGrid, nextGrid); 85 | currGrid = nextGrid; % Move to the next grid 86 | % 2.2.4 Update J_k(r) and corresponding localDelta to prepare for the next move 87 | toBeVisited(currGrid) = 0; 88 | for i = 1:size(localDelta, 1) 89 | if toBeVisited(i) == 0 90 | localDelta(currGrid, i) = 0; 91 | localDelta(i, currGrid) = 0; 92 | end 93 | end 94 | indexDelta = indexing(localDelta, currGrid, toBeVisited); 95 | numAvailable = length(indexDelta); 96 | counter = counter + 1; 97 | end % the ant reaches to the destination and stop the loop 98 | % 2.3 Record the job done by the ant 99 | pathStorage{Gen, Ant} = pathRecord; % Save the path 100 | if pathRecord(end) == destination 101 | if pathLength(Gen, Ant) < minTourLength % Does the ant find a shorter path? 102 | minTourLength = pathLength(Gen, Ant); % Yes, save the new shortest path 103 | bestGen = Gen; % Remember the generation that achieved this 104 | bestAnt = Ant; % Remember the ant that achieved this 105 | end 106 | else % When the ant is not successful, offset its length in this iteration 107 | pathLength(Gen, Ant) = 0; 108 | end 109 | end % The ant finishes the forage this time 110 | % 2.4 Update the intensity of pheromone on the path 111 | deltaTau = zeros(size(tau)); 112 | for i = 1: numAnts 113 | if pathLength(Gen, Ant) 114 | pathTemp = pathStorage{Gen, Ant}; 115 | pathInterval = length(pathTemp) - 1; 116 | for j = 1: pathInterval 117 | deltaTau(pathTemp(j), pathTemp(j + 1)) = 1 / pathLength(Gen, Ant); 118 | deltaTau(pathTemp(j + 1), pathTemp(j)) = 1 / pathLength(Gen, Ant); 119 | end 120 | end 121 | end 122 | tau = (1 - alpha) .* tau + deltaTau; % pheromone evaporation rule 123 | end % All the ants in this generation finished forage; restart a new generation now 124 | 125 | % 3. VISUALISATION 126 | showPath(gridMap,SIZE, pathStorage, bestGen, bestAnt); 127 | showLength(pathLength); 128 | -------------------------------------------------------------------------------- /2-LineObstacles/ConvertXY.m: -------------------------------------------------------------------------------- 1 | function [x, y] = ConvertXY(gridNum, rowMap) % Extract x coordinate from the serial number of the location 2 | x = ceil(gridNum / rowMap) - 0.5; 3 | y = rowMap - mod(gridNum, rowMap) + 0.5; 4 | if y == rowMap + 0.5 5 | y = 0.5; 6 | end 7 | end 8 | 9 | -------------------------------------------------------------------------------- /2-LineObstacles/Iterations.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/Iterations.jpg -------------------------------------------------------------------------------- /2-LineObstacles/Log.md: -------------------------------------------------------------------------------- 1 | This is the 2nd version. 2 | **Date: 20 March 2022 3 | 4 | # Actions Log 5 | 6 | At this stage, the main task is to **test different scenarios** and see if there is any bugs prompt out. 7 | 8 | Then, modify the programme to improve the performance in terms of **accuracy, stability and response speed**. 9 | 10 | Lastly, prepare **presentation** to introduce algorithms used in detail at Monday's meeting. 11 | 12 | ## 1. Obstacles 13 | 14 | ### Obstacles added. 15 | ``` 16 | gridMap(2:9, 5) = 1; 17 | ``` 18 | * Think about: 19 | 20 | * How to find the best solution by hand-calculation and compare with the result from the programme when the map gets more and more complex? 21 | * TBC... 22 | 23 | ## 2. Δ(r, s) array generation 24 | 25 | ### 🐭 Bug occurred: 26 | 27 | * Can't detect obstacle at gridMap(88) and some surrounding grids when 28 | ``` 29 | gridMap(9, 4:8) = 1; 30 | gridMap(4:9, 9) = 1; 31 | ``` 32 | * Check data detail in file [globalDelta_Bug.xlsx](../1-Non-Obstacles/globalDelta_Bug.xlsx) 33 | * Original code in [delta_r2s.m](../1-Non-Obstacles/delta_r2s.m) 34 | ``` 35 | function D = delta_r2s(arr) % delta_(r,s): cost from grid r to grid s 36 | % r and s must be next to each other ssss 37 | row = size(arr, 1); % the number of rows of the array 38 | column = size(arr, 2); 39 | D = zeros(row*column); % initialise the delta_(r,s) matrix 40 | % Outer Loop 41 | for i = 1:row 42 | for j = 1:column % Outter Loop: Traverse all the grids to set their cost 43 | % of traveling to the local surrouding grids 44 | if arr(i, j) == 0 % grid r must not be a obstacle 45 | 46 | % Inner Loop 47 | for x = 1:row 48 | for y = 1:column % Inner Loop: find the surrounding available 49 | % grids s and return the cost value 50 | % (distance) to the exact variable in 51 | % the matrix 52 | if arr(x, y) == 0 % the surrounding grid must be clear 53 | rel_x = abs(i - x); % relative distance between grid r and s on x direction 54 | rel_y = abs(j - y); % relative distance between grid r and s on y direction 55 | if rel_x + rel_y == 1 || (rel_x == 1 && rel_y == 1) 56 | % When this grid is one of the adjacent 57 | % grids in 8 directions 58 | D((i-1)*column + j, (x-1)*column + y) = sqrt(rel_x^2 + rel_y^2); 59 | % D(r, s), r and s are indices of one-dimensional matrix 60 | % Return relative distance between r and s 61 | end 62 | end 63 | end 64 | end 65 | end 66 | end 67 | end 68 | 69 | end 70 | ``` 71 | 72 | ### ✔️ Solution: 73 | 74 | * Updated the generation process [calcDelta.m](calcDelta.m) 75 | 76 | ``` 77 | function D = calcDelta(arr) 78 | row = size(arr, 1); % the number of rows of the array 79 | column = size(arr, 2); 80 | D = zeros(row*column); % initialise the delta_(r,s) matrix 81 | for i = 1: row*column 82 | if arr(i) == 0 83 | for j = 1: row*column 84 | if arr(j) == 0 && (abs(i - j) == 11 || abs(i - j) == 10 || abs(i - j) == 9 || abs(i - j) == 1) 85 | [x1, y1] = ConvertXY(j, row); 86 | [x2, y2] = ConvertXY(i, row); 87 | D(i, j) = sqrt((x1 - x2)^2 + (y1 - y2)^2); 88 | D(j, i) = sqrt((x1 - x2)^2 + (y1 - y2)^2); 89 | end 90 | end 91 | end 92 | end 93 | 94 | end 95 | ``` 96 | 97 | ## 3. Modify pheromone evaporation rate Δτ 98 | 99 | Changing _Δτ_ can improve the accuracy, which means the higher chance of returning the expected answer. 100 | 101 | 102 | 103 | ### 🐭 Bug: Path length converging at the highest level 104 | 105 | The length-iteration diagram was expected to show the convergence at the lowest value instead of the greatest value. 106 | 107 | ![shortest path](robotPathRecords.jpg) 108 | |:--:| 109 | | *Robot path records* | 110 | 111 | ![diagram](Iterations.jpg) 112 | |:--:| 113 | | *Length-Iteration diagram* | 114 | 115 | ### 📆 21 March 2022 Update 116 | 117 | * Pheromone update rules mis-coded ❎ 118 | ``` 119 | % 2.4 Update the intensity of pheromone on the path 120 | deltaTau = zeros(size(tau)); 121 | for i = 1: numAnts 122 | if pathLength(Gen, Ant) 123 | pathTemp = pathStorage{Gen, Ant}; 124 | pathInterval = length(pathTemp) - 1; 125 | for j = 1: pathInterval 126 | deltaTau(pathTemp(j), pathTemp(j + 1)) = 1 / pathLength(Gen, Ant); 127 | deltaTau(pathTemp(j + 1), pathTemp(j)) = 1 / pathLength(Gen, Ant); 128 | end 129 | end 130 | end 131 | tau = (1 - alpha) .* tau + deltaTau; % pheromone evaporation rule 132 | ``` 133 | #### :hammer: Debug Attempt: 134 | 135 | the _pathLength(Gen, **Ant**)_ should be _pathLength(Gen, **i**)_ 136 | 137 | :black_nib: Thougth: If it followed the first code, it would only update the pheromone based on **the last ant's path**. 138 | 139 | * Code: 140 | ``` 141 | % 2.4 Update the intensity of pheromone on the path 142 | deltaTau = zeros(size(tau)); 143 | for i = 1: numAnts 144 | if pathLength(Gen, i) 145 | pathTemp = pathStorage{Gen, i}; 146 | for j = 1: size(pathTemp) 147 | deltaTau(pathTemp(j)) = deltaTau(pathTemp(j)) + 1 / pathLength(Gen, i); 148 | end 149 | end 150 | end 151 | tau = (1 - alpha) .* tau + deltaTau; % pheromone evaporation rule 152 | ``` 153 | * Resulte: Still not converging to the smallest value 154 | * Change the Pheromone update rule τ(r, s), using 2D array 155 | 156 | ![shortest path](Results/robotPath_newPheromoneupdate.jpg) 157 | |:--:| 158 | | *Robot path records: changed the pheromone update rule* | 159 | 160 | ![diagram](Results/Length-Iteration_newPheromoneupdate.jpg) 161 | |:--:| 162 | | *Length-Iteration diagram: changed the pheromone update rule* | 163 | 164 | * Change the relative importance of pheromone = 4 165 | ``` 166 | % Original 167 | stateTransProb(i) = tau(indexDelta(i)) * eta(indexDelta(i))^beta; 168 | 169 | % NEW 170 | stateTransProb(i) = tau(indexDelta(i))^4 * eta(indexDelta(i))^beta; 171 | ``` 172 | ![shortest path](Results/robotPath_relativeImportance4.jpg) 173 | |:--:| 174 | | *Robot path records: changed the pheromone update rule* | 175 | 176 | ![diagram](Results/Length-Iteration__relativeImportance4.jpg) 177 | |:--:| 178 | | *Length-Iteration diagram: changed the pheromone update rule* | 179 | -------------------------------------------------------------------------------- /2-LineObstacles/Results/Length-Iteration__relativeImportance4.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/Results/Length-Iteration__relativeImportance4.jpg -------------------------------------------------------------------------------- /2-LineObstacles/Results/Length-Iteration_newPheromoneupdate.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/Results/Length-Iteration_newPheromoneupdate.jpg -------------------------------------------------------------------------------- /2-LineObstacles/Results/robotPath_newPheromoneupdate.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/Results/robotPath_newPheromoneupdate.jpg -------------------------------------------------------------------------------- /2-LineObstacles/Results/robotPath_relativeImportance4.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/Results/robotPath_relativeImportance4.jpg -------------------------------------------------------------------------------- /2-LineObstacles/calcDelta.m: -------------------------------------------------------------------------------- 1 | function D = calcDelta(arr) 2 | row = size(arr, 1); % the number of rows of the array 3 | column = size(arr, 2); 4 | D = zeros(row*column); % initialise the delta_(r,s) matrix 5 | for i = 1: row*column 6 | if arr(i) == 0 7 | for j = 1: row*column 8 | if arr(j) == 0 && (abs(i - j) == 11 || abs(i - j) == 10 || abs(i - j) == 9 || abs(i - j) == 1) 9 | [x1, y1] = ConvertXY(j, row); 10 | [x2, y2] = ConvertXY(i, row); 11 | D(i, j) = sqrt((x1 - x2)^2 + (y1 - y2)^2); 12 | D(j, i) = sqrt((x1 - x2)^2 + (y1 - y2)^2); 13 | end 14 | end 15 | end 16 | end 17 | 18 | end 19 | 20 | -------------------------------------------------------------------------------- /2-LineObstacles/calcHeuristic.m: -------------------------------------------------------------------------------- 1 | function Eta = calcHeuristic(arr, grid) 2 | dim1 = size(arr, 1); 3 | dim2 = size(arr, 1) * size(arr, 1); 4 | for i = 1:dim2 5 | [desti_x, desti_y] = ConvertXY(grid, dim1); 6 | [x, y] = ConvertXY(i, dim1); 7 | distance = sqrt((x - desti_x)^2 + (y - desti_y)^2); 8 | if i ~= grid 9 | arr(i) = 1 / distance; 10 | else 11 | arr(grid) = 100; 12 | end 13 | end 14 | Eta = arr; 15 | end 16 | 17 | -------------------------------------------------------------------------------- /2-LineObstacles/indexing.m: -------------------------------------------------------------------------------- 1 | function index = indexing(delta, grid, J_kr) % Return the indices of available grids 2 | index = delta(grid, :); % Store the cost required to the next gird 3 | arr = find(index); % Scope down available adjacent grids indices; store the complete indices 4 | % Double check for visited grids to avoid self-repeating 5 | for i = 1: length(arr) 6 | if J_kr(arr(i)) == 0 7 | index(arr(i)) = 0; 8 | end 9 | % What about excluding obstacles? grids == 1? 10 | % Why is the path going straight down? while heuristic value are 11 | end 12 | index = find(index); 13 | end 14 | 15 | -------------------------------------------------------------------------------- /2-LineObstacles/robotPathRecords.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/2-LineObstacles/robotPathRecords.jpg -------------------------------------------------------------------------------- /2-LineObstacles/showLength.m: -------------------------------------------------------------------------------- 1 | function showLength(arr) 2 | figure(2) 3 | 4 | row = size(arr, 1); 5 | column = size(arr, 2); 6 | minLength = zeros(row, 1); 7 | for i = 1: row 8 | arrRows = arr(i, :); 9 | index = arrRows ~= 0; 10 | arrRows = arrRows(index); 11 | minLength(i) = min(arrRows); 12 | end 13 | plot(minLength); 14 | hold on 15 | grid on 16 | title('Path Length Changes During Iterations'); 17 | xlabel('Iteration'); 18 | ylabel('Minimum Path Length'); 19 | end 20 | 21 | -------------------------------------------------------------------------------- /2-LineObstacles/showPath.m: -------------------------------------------------------------------------------- 1 | function showPath(map, size, path, gen, ant) 2 | figure(1); 3 | for i = 1: size*size 4 | if map(i) == 1 5 | [x_p, y_p] = ConvertXY(i, size); 6 | x1 = x_p - 0.5; y1 = y_p + 0.5; 7 | x4 = x_p - 0.5; y4 = y_p - 0.5; 8 | x2 = x_p + 0.5; y2 = y_p + 0.5; 9 | x3 = x_p + 0.5; y3 = y_p - 0.5; 10 | fill([x1, x2, x3, x4], [y1, y2, y3, y4], [0.2 0.2 0.2]); 11 | hold on; 12 | else 13 | [x_p, y_p] = ConvertXY(i, size); 14 | x1 = x_p - 0.5; y1 = y_p + 0.5; 15 | x4 = x_p - 0.5; y4 = y_p - 0.5; 16 | x2 = x_p + 0.5; y2 = y_p + 0.5; 17 | x3 = x_p + 0.5; y3 = y_p - 0.5; 18 | fill([x1, x2, x3, x4], [y1, y2, y3, y4], 'w'); 19 | hold on; 20 | end 21 | end 22 | hold on; 23 | title('Map'); 24 | xlabel('x-axis'); 25 | ylabel('y-axis'); 26 | pathRecord = path{gen, ant}; 27 | ctr = length(pathRecord); 28 | robot_x = zeros(ctr, 1); 29 | robot_y = zeros(ctr, 1); 30 | for j = 1: ctr 31 | [robot_x(j), robot_y(j)] = ConvertXY(pathRecord(j), size); 32 | end 33 | plot(robot_x, robot_y); 34 | end 35 | 36 | -------------------------------------------------------------------------------- /3 - Convergence/Log.md: -------------------------------------------------------------------------------- 1 | # Changelog 2 | 3 | This week, I should finalise the experiment model. 4 | 5 | ## 🔖 To-do list updated on [25-03-2022] 6 | 7 | * Learn from the working example code 8 | * ✔️ Resolve the flawed condition on line 8 in ```calcDelta``` function 9 | * ✔️Inspect pheromone update process and foraging process 10 | * ✔️Would be helpful to check ```calcDelta``` ```ConvertXY``` ```tau``` ```deltaTau``` ```pathRecord``` ```globalDelta``` ```pathLength``` 11 | * ✔️ Should expect the converge result in the end 12 | * Analyse the map size, iterations, number of ants, relative importance, pheromone matrix to solve the convergence issue 13 | * Maybe it is not supposed to converge at the minimal value? just close 14 | 15 | ## 📍 [Unreleased] 16 | 17 | ### Parameters modifications 18 | 19 | * Q - pheromone intensity coefficient can be considered later 20 | 21 | ### Visualisation 22 | 23 | * Pheromone intensity map - distinguish each level with a set of colours 24 | * Maybe show it changing along the iterations 25 | 26 | ### Compare 'Without pheromone' & 'With Pheromone' 27 | 28 | * 📉 Shown difference on line chart: 29 | * Pure lucky & guided by colony's trend 30 | 31 | ## 📖 Week 8 Meeting [21-03-2022] - Outlines 32 | * The ants should converge to the optimal solution in the end. 33 | * Try visualise the process: from arbitrarily-generated path to the pheromone-guided-generated path. 34 | * Try visualise pheromone change progress 35 | * Compare results from different sets of parameters: the number of ants, the number of iterations and pheromone evaporation rate. 36 | * Mann-whitney test 37 | * Imitate the excellent example sent by Dr Castellani 38 | 39 | ## 🆕 Update [25/03/2022] 40 | 41 | Focused on pheromone update code. 42 | 43 | ### 🔴 Issues 44 | 45 | * ✔️ Collision: In rare cases, robot still collides with obstacles. 46 | * Instead of collision, it is actually flying over to distanced grids because of wrong cost measure **δ** set up. 47 | 48 | 49 | |:--:| 50 | | *Figure 1* | 51 | 52 | When I changed the code to... (Though that's not the key point) 53 | ``` 54 | tau = ones(SIZE); 55 | tau = 10.*tau; % Set the initial pheromone of all edges (r,s) as 8 56 | alpha = 0.4; % the Evaporation rate of pheromone 57 | ``` 58 | 59 | * Pheromone matrix **τ** updates abnormally 60 | * Only grid 1 was updated when other grids in **Δτ** remain zero 61 | 62 | 63 | ### 👨‍🔧 Actions 64 | 65 | * Checked ``` tau ``` and ``` deltaTau ``` data 66 | * Only grid 1 in ``` deltaTau ``` has value; Grid 1 in ``` tau ``` was 16.2298 67 | * Why is only grid 1 updated while the entire path should be passed new values? 68 | 69 | 70 | * Changed line 118 from ``` for j = 1: size(pathTemp)``` to ``` for j = 1: length(pathTemp) ``` 71 | * pheromone matrix **τ** seems normal now 72 | 73 | * Collision issue 74 | * Checked ```pathRecord``` to see the path trail and ```globalDelta``` to see the available grids 75 | * Checked ```calDelta``` function: the condition logic is flawe 76 | 77 | calDelta function line 8: 78 | ``` 79 | if arr(j) == 0 && (abs(i - j) == 11 || abs(i - j) == 10 || abs(i - j) == 9 || abs(i - j) == 1) 80 | ``` 81 | ❗ abs(i - j) == 10 and abs(i - j) == would recognise girds at the other end of the same column by mistake. 82 | 83 | ## 🆕 Update [26-03-2022] 84 | 85 | Focused on fixing the **condition** of scoping down available surrounding grids in cost measure **δ(r, s)** function 86 | 87 | ### 🔴 Issues 88 | 89 | * Fluctuation occurs occasionally 90 | 91 | Parameters setting 92 | ``` 93 | SIZE = 10; 94 | gridMap = zeros(SIZE); % Environment configuration: 0 for clear path, 1 for obstacles 95 | gridMap(5, 2:9) = 1; 96 | gridMap(9, 4:8) = 1; 97 | gridMap(2:9, 5) = 1; 98 | numAnts = 5; % the number of ants in a colony 99 | numGen = 200; % the number of generations (iterations) 100 | ``` 101 | 102 | 103 | |:--:| 104 | | *Figure 2* | 105 | 106 | 107 | |:--:| 108 | | *Figure 3* | 109 | 110 | 111 | When ```numAnts = 10; numGen = 200; ```: 112 | 113 | |:--:| 114 | | *Figure 4* | 115 | 116 | * Not converge to the minimal path length when increasing the iterations 117 | 118 | ```numAnts = 50; numGen = 100; alpha = 0.4``` 119 | ![issueConverge](https://user-images.githubusercontent.com/69563490/160237959-503e4642-ba08-4c3f-a278-977c8cc69b9d.jpg) 120 | 121 | ```numAnts = 50; numGen = 200; ``` 122 | ![issueNotConverge](https://user-images.githubusercontent.com/69563490/160237977-ffde265e-aa4d-49d5-bd44-2d688664bc09.jpg) 123 | 124 | * Fluctuate when relative importance of pheromone are less 125 | Could not converge 126 | ``` 127 | numAnts = 10; numGen = 100; alpha = 0.4; beta = 7; 128 | ... 129 | stateTransProb(i) = tau(indexDelta(i)) * eta(indexDelta(i))^beta; 130 | ``` 131 | * Potential sources: 132 | Pheromone evaporation rule; 133 | 134 | 135 | ### 👨‍🔧 Actions 136 | 137 | * Modified conditions (line 8) of recognising surrounding grids in function ```calcDelta(arr)``` 138 | * Before update: 139 | ``` 140 | if arr(j) == 0 && (abs(i - j) == 11 || abs(i - j) == 10 || abs(i - j) == 9 || abs(i - j) == 1) 141 | ``` 142 | Such condition will not only select out surrounding grids, but also grids on the both end of each column, causing issues of fake collision (or flying to the other end) shown in _Figure 1_ in **Issue [25-03-2022]** section. 143 | 144 | * After update: 145 | ``` 146 | if arr(j) == 0 && (abs(i - j) == 11 || abs(i - j) == 10 || abs(i - j) == 9 || abs(i - j) == 1) 147 | [xi, yi] = ConvertXY(i, row); 148 | [xj, yj] = ConvertXY(j, row); 149 | dis = sqrt((xi - xj)^2 + (yi - yj)^2); % Check the distance 150 | if dis <= sqrt(2) % Make sure it's surrouding the current grid 151 | D(i, j) = dis; % sqrt((x1 - x2)^2 + (y1 - y2)^2); 152 | D(j, i) = dis; % sqrt((x1 - x2)^2 + (y1 - y2)^2); 153 | end 154 | end 155 | ``` 156 | * Updated results 157 | 158 | Surprisingly, the convergence issue has been solved after solving the fake collision issue. 159 | 160 | ``` numAnts = 20; numGen = 50;``` 161 | 162 | |:--:| 163 | | *Figure 5* | 164 | 165 | 166 | |:--:| 167 | | *Figure 6* | 168 | 169 | ``` numAnts = 50; numGen = 100; ``` 170 | 171 | |:--:| 172 | | *Figure 7* | 173 | 174 | * Changed relative importance 175 | 176 | Works better 177 | ``` 178 | numAnts = 10; numGen = 100; alpha = 0.4; beta = 7; 179 | ... 180 | stateTransProb(i) = tau(indexDelta(i))^5 * eta(indexDelta(i))^beta; 181 | ``` 182 | Result: 183 | 184 | 185 | |:--:| 186 | | *Figure 8* | 187 | 188 | Could not converge 189 | ``` 190 | numAnts = 10; numGen = 100; alpha = 0.4; beta = 7; 191 | ... 192 | stateTransProb(i) = tau(indexDelta(i)) * eta(indexDelta(i))^beta; 193 | ``` 194 | Result: 195 | 196 | 197 | |:--:| 198 | | *Figure 9* | 199 | 200 | * Modified relative importance making pheromone more important than heuristic value 201 | * It can converge now 202 | * How does the example work with heuristic value more important than pheromone?? 203 | 204 | 205 | ### Questions 206 | 207 | * 报告花多少时间 208 | * 有些情况会出问题：我代码写的和别人的不太一样，得到的结果有出入，但符合预期，要怎么办 209 | * 怎样写好报告 210 | * 本科毕设需要提出创新的点吗？从什么角度入手比较好 211 | -------------------------------------------------------------------------------- /4 - Final Stage/Log.md: -------------------------------------------------------------------------------- 1 | # ChangeLog 2 | 3 | At this stage, the main focus is on thinking, questioning, analysing and concluding the work that has been done. 4 | 5 | ## Report Writing 6 | 7 | ### General instruction 8 | 9 | * From the [blog](https://academia.stackexchange.com/questions/76843/when-writing-a-paper-whats-the-difference-between-contributions-and-objectives) 10 | 11 | "Usually, the scheme of a scientific study has a hourglass shape scheme, from wide, to narrow, and wide again. 12 | 13 | In the first wide, we use the introduction to describe the general topography and bring focus to a particular problem. Then we narrow down to describe how our experiment is designed, data are collected, analysis is done, etc. 14 | 15 | When results are ready, we interpret them, and then again widen up the scope to talk about implications, applications, and recommendations in the Discussion section. 16 | 17 | In my opinion, objectives are a list of actions to perform, or questions to answer so that you can complete the narrow part of the study scheme. Contributions are a list of applications or significance to other researchers that brought about by your work, and they are perhaps more fitting in the second wider part of the study scheme." 18 | 19 | 20 | ## Meeting with Feiying 21 | 22 | **Takeaways:** 23 | 24 | * 精英策略 25 | * Error code 误差曲线 26 | * 准确率 27 | * 用了多少次迭代 28 | * 讨论 29 | * 为什么用这个方法好 - 如何更有效率，更完善 - 探讨指标 30 | * 为什么这样会变好 - 不消耗计算负担的参数，提升性能 - 有什么参数可以实现这个 31 | * 以辩证角度评论文献综述，逻辑连贯要支撑我为什么要做这个项目 32 | * 思考 33 | * 一定要收敛吗？这是一个概率选择的过程，总会出现偏差。为什么会出现波动？说明信息素和启发素浓度不够？ 34 | * 每次迭代取得最优解的蚂蚁的数量占比的变化有没有分析价值？ 35 | * 能否展示每次迭代所有蚂蚁的路径在同一个地图并随着迭代次数的变化情况来看信息素有没有发挥的作用？ 36 | * 改变了相对的重要性参数 37 | 38 | ## 📍 [Unreleased] 39 | 40 | ### 🚩 Elitism / Elitist Preservation 41 | 42 | Proposed by Kenneth A. De Jong from his Ph.D dissertation on Genetic Algorithms 43 | 44 | #### Definition 45 | 46 | maintain the best solution found over time before selection 47 | 48 | #### Potential Application 49 | 50 | Apply it to ACO ensuring that the algorithm converge at the best solution 51 | 52 | Refer to this [blog about Elitist Preservation](https://www.cnblogs.com/devilmaycry812839668/p/6445762.html) for the details of the method 53 | 54 | [Python example](https://dothinking.github.io/2018-10-27-%E9%81%97%E4%BC%A0%E7%AE%97%E6%B3%95%EF%BC%9A%E6%94%B9%E8%BF%9B%E6%96%B9%E5%90%91%E4%B9%8B%E7%B2%BE%E8%8B%B1%E7%AD%96%E7%95%A5/#:~:text=%E8%87%AA%E9%80%82%E5%BA%94%E7%AD%96%E7%95%A5-,%E7%B2%BE%E8%8B%B1%E4%BF%9D%E7%95%99%E7%AD%96%E7%95%A5,%E4%BF%9D%E7%95%99(Elitist%20Preservation)%20%E7%AD%96%E7%95%A5%E3%80%82) 55 | 56 | ## 🚩 57 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 |
7 |
8 | 9 | Logo 10 | 11 | 12 |

Ant Colony Optimisation (ACO) for Robot Path Planning

13 | 14 |

15 | This report presents the visualisations and analysis of using ACO to address NP-hard complexity within Robot Path Planning problems. 16 |
17 | Explore the docs » 18 |
19 |
20 | View Demo 21 | · 22 | Report Bug 23 | · 24 | Request Feature 25 |

26 |
27 | 28 | 29 | 30 | 31 |
32 | Table of Contents 33 |
    34 |
  1. 35 | About The Project 36 | 39 |
    2. 40 |
  2. 41 | Getting Started 42 | 46 |
    3. 47 |
  3. Usage
    4. 48 |
  4. Roadmap
    5. 49 |
  5. Contributing
    6. 50 |
  6. License
    7. 51 |
  7. Contact
    8. 52 |
  8. Acknowledgments
    9. 53 |
54 |
55 | 56 | matlab 57 | 58 | 59 | ## About The Project 60 | 61 | **Why robot path planning matters**: Picture a robot in your local warehouse, deftly weaving through stacks of boxes to fulfill your online shopping orders. Or imagine self-driving cars effortlessly navigating complex city streets. It's all about efficient path planning! 📦🚗 62 | 63 | Robot path planning is like a GPS for robots. It helps them find the best way from A to B while dodging obstacles, saving time, and conserving energy. It's the brain behind the bots! 64 | 65 | **Enter Ant Colony Optimisation!** Now, here's where it gets exciting. We're introducing ACO, a nature-inspired algorithm that mimics how ants find the shortest path to food. 🐜🍔 66 | 67 | ACO is our robot's secret sauce. It's like giving them a sixth sense for navigating tricky terrains. By laying down 'virtual pheromones,' our robots learn and adapt, discovering the optimal routes in changing environments. 68 | 69 | ### A glimpse at the artificial ant colony 70 | 71 | Take a look at the final outcome of the work! 72 | 73 | ![](image/Demo.gif) 74 | 75 | * 🎞️ Above animation visualises how the **artificial ant colony** navigate in a world full of obstacles. 76 | * 🧀 Their goal is to identify the best (shortest and obstacle-free) route from their nest 🟢 to the food resources 🔴. 77 | * ⛵ **Exploration**: They firstly bravely explore many possible routes because of no prior knowledge of the world. 78 | * 🧔 **Learn from excellent peers**: Then they quickly learn from their past experiences. 79 | * 🥇 **Team work wins!** Eventually, all of them merge on the optimal route! 80 | 81 |

(back to top)

82 | 83 | ### How it works 84 | 85 | - :ant: Ant Colony Optimization (ACO) is like a picnic adventure for ants! 86 | - :sandwich: Ants need to find the shortest path to their favorite sandwich. 87 | - :world_map: Many paths to choose from, no GPS for ants! 88 | - :walking_man: Ants leave scented breadcrumbs called **'pheromones'** as they explore. 89 | - :nose: Strong pheromone smell means "This way to the sandwich!" 90 | - :walking_woman: Other ants **follow the strong pheromone trail**. 91 | - :briefcase: More ants travel the same path, making the trail even stronger. 92 | - :railway_track: Ants **converge** on the shortest, most efficient route. 93 | - :robot: ACO mimics this natural process to solve complex problems for computers. 94 | - :exploding_head: Helps robots find their way through mazes and optimises delivery routes. 95 | - :earth_americas: Nature's way of showing us how to navigate the world, one picnic at a time! :ant::sandwich::world_map: 96 | 97 | ## Getting started 98 | 99 | 100 | 101 | ### Theory 102 | 103 | 104 | 105 | Extensions: 106 | 107 | * Ant Colony System 108 | * Elitist Ant System 109 | * Max-min Ant System 110 | 111 | ### Mapping envrionment 112 | 113 | Grid mapping method--discretisation of real world scenarios 114 | 115 | 116 | 117 | ### Hyper-parameters optimisation 118 | 119 | Its performance and rate of convergence are sensitive to the chosen parameter values, and especially to the value of the pheromone evaporation rate. 120 | 121 | 122 | 123 | ### Test and results 124 | 125 | **Test Map** 126 | 127 | 128 | 129 | 130 | 131 | **Results** 132 | 133 | 134 | 135 | 136 | 137 | ## Performance analysis 138 | 139 | 140 | 141 | ## Archived Logs 142 | 143 | The following is archived records of edit history. 144 | 145 | ### Update dates & details 146 | Ver. 17 Mar 2022 147 | * Bug cleared 148 | * No obstacles in the map 149 | ``` 150 | numAnts = 17; 151 | numGen = 20; 152 | ``` 153 | * To ensure to find the diagonal-shortest path, deployed precedent parameters 154 | 155 | Ver. 20 Mar 2022 156 | * Added obstacles to the map 157 | ``` 158 | gripMap(2:9, 5) = 1; 159 | ``` 160 | * Ran well, but need to improve the accuracy and stability 161 | 162 | * Solved a fail-to-detect-obstacles issue 163 | * When using the Delta_r2r(arr) function, globalDelta stored obstacles in the array by mistake. 164 | * Bug source still not found 165 | 166 | 167 | 168 | [JQuery-url]: https://jquery.com 169 | -------------------------------------------------------------------------------- /image/ACO Flowchart.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/ACO Flowchart.png -------------------------------------------------------------------------------- /image/ACOProjectLogo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/ACOProjectLogo.png -------------------------------------------------------------------------------- /image/Demo.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Demo.gif -------------------------------------------------------------------------------- /image/Map-Algorithm Boxchart.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Map-Algorithm Boxchart.jpg -------------------------------------------------------------------------------- /image/Map3.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Map3.jpg -------------------------------------------------------------------------------- /image/Map_Easy.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Map_Easy.png -------------------------------------------------------------------------------- /image/Map_Hard.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Map_Hard.png -------------------------------------------------------------------------------- /image/Min Path Length-Data (1).jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Min Path Length-Data (1).jpg -------------------------------------------------------------------------------- /image/Min path length - data.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Min path length - data.jpg -------------------------------------------------------------------------------- /image/ParaSelec_ACS_Mark.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/ParaSelec_ACS_Mark.jpg -------------------------------------------------------------------------------- /image/ParaSelec_MMAS_1.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/ParaSelec_MMAS_1.jpg -------------------------------------------------------------------------------- /image/ParaSelec_eAS_1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/ParaSelec_eAS_1.png -------------------------------------------------------------------------------- /image/PathPlanningAlgorithmsOverview.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/PathPlanningAlgorithmsOverview.png -------------------------------------------------------------------------------- /image/PerformanceAnalysis_CS.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/PerformanceAnalysis_CS.png -------------------------------------------------------------------------------- /image/PerformanceAnalysis_Runtime.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/PerformanceAnalysis_Runtime.png -------------------------------------------------------------------------------- /image/Route Map (1).jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Route Map (1).jpg -------------------------------------------------------------------------------- /image/Route Map (2).jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Route Map (2).jpg -------------------------------------------------------------------------------- /image/Route Map.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/Route Map.jpg -------------------------------------------------------------------------------- /image/aco_diagram.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/aco_diagram.jpg -------------------------------------------------------------------------------- /image/converge_map1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/converge_map1.png -------------------------------------------------------------------------------- /image/converge_map2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/converge_map2.png -------------------------------------------------------------------------------- /image/final_demo.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/final_demo.gif -------------------------------------------------------------------------------- /image/image_log.md: -------------------------------------------------------------------------------- 1 | All the images are stored here. 2 | -------------------------------------------------------------------------------- /image/matlab.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/matlab.png -------------------------------------------------------------------------------- /image/route_map1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/route_map1.png -------------------------------------------------------------------------------- /image/route_map2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/onlyEugeneLi/ACO-RobotPathPlanning-MATLAB/cb5c0c694d91860b83171b91530f74c3ea0f1ac4/image/route_map2.png --------------------------------------------------------------------------------