Enhanced Artificial Bees Colony Algorithm for Robot Path Planning

Transcription

1 Enhanced Artificial Bees Colony Algorithm for Robot Path Planning Department of Computer Science and Engineering, Jaypee Institute of Information Technology, Noida ABSTRACT: This paper presents an enhanced algorithm for planning the robot path using Artificial Bee Colony Algorithm. This algorithm is used to find optimal and collision free path for a predefined starting and ending point in a given environment. The ABC algorithm used is inspired by the collective behavior of bees to find better food sources around the hive. Keywords: Robot path planning, Artificial Bees Colony Algorithm, Swarm Intelligence, Collision Avoidance. [1] INTRODUCTION Robotics is emerging as a powerful, rapidly growing technique which deals with automation of machines that can help the humans by taking their place in manufacturing processes or dangerous environments or resemble humans in behaviour, appearance and/or cognition. The field of robotics has resulted in the emergence of bio inspired robotics due to the inspiration from nature. The research process for robot path planning started in the middle of 1960s. The research accelerated after the release of publication of Lozano-Pe rez [1] in After 34 years of research, approaches for robot path planning can be divided into two major categories: classical and heuristics. The potential field methods [2], visibility graph methods [3] and grid methods are famous classical methods for robot planning. These approaches suffer from various limitations such as trapping in local minima, obstacles are so closely spaced that there is no passage, lower search space and many more. The NP completeness nature of path planning problem persuaded the researchers to switch from classical approaches to heuristic approaches. Some of the heuristics approaches applied to path planning problem are Fuzzy Logic [4], neural networks [7], genetic algorithm [8] and ant colony optimisation [9]. One of the important issue in the field of robotics is that of path planning.. It focusses on finding the optimal path from one location to another with an ultimate objective of 71

2 Enhanced Artificial Bees Colony Algorithm for Robot Path Planning minimizing the path length. In other words, it can be mapped to the problem of shortest path problem between any two nodes in a graph. Generally, there are many paths for a robot to reach the target but, in fact the best path is selected on basis of some guidelines such as path which is of shortest length, energy consumption is least or minimum time is taken to travel the path.[5][6] The path planning problem can be solved in two ways: off line and on line. In first case, the environment and the positions of obstacles are known beforehand in determining the optimal collision free path for the robot. In second case, the information about the environment and the obstacles is obtained with the help of some sensing device. Section 2 deals with the natural bees, their foraging behaviour and the core ideas of the Bees algorithm. Section 3 describes the proposed methodology for solving robot path planning. Section 4 describes the simulation results and Section 5 deals with the results and conclusion. [2] Honey Bees in Nature In order to find more number of food sources, colony of honey bees is extended in multiple directions for long distances at a same time. [10, 11].The prosperity of colony is due to the deployment of its foragers to good fields. They follow the principle that flower patches with plenty amount of nectar is visited by more bees and the ones having less amount of nectars have fewer visitors [12, 13]. The food foraging behaviour involving intelligence of honey bees is the main motivation for Artificial Bee Colony Algorithm. One of the most intelligent swarms existing in nature is honey bees swarm which follows collective intelligent method during the food search. The various qualities possessed by honey bee swarms are communicating of information by the bees, memorizing of the environment, storing and sharing the information and taking decisions based on the information. According to the changing environment, the updation of swarm by itself takes place. Then tasks are assigned dynamically and further movement occurs due to social learning and teaching. This behaviour of bees having intelligence forms the motivation for researchers to simulate above foraging behaviour of the bee swarm. The three major elements as proposed by D. Karaboga [14] are: food sources, employed and unemployed foragers. The employed bees are associated with food sources which are appropriate. The knowledge about the food sources is present with the employed bees. Employed bees are responsible for the exploitation of food sources. The employed bee changes to unemployed when the food sources become abandoned. The bees which do not have any information about the food source and searches for a food source for exploiting are unemployed foragers. The unemployed bees are classified into two types: scout bees and onlooker bees. Scout bees are those unemployed bees that search for new food sources randomly which surrounds the hive. Onlooker bees are those bees which uses the information conveyed by the waggle dance to search for new food sources for exploitation. The third element is the food source rich in nectar and present close to hive. Thus number of solutions is represented by the number of food sources in

3 the algorithm. Further, the position of a favourable solution is represented by the location of food sources for optimization problem because the fitness cost of the solution corresponds to the nectar quality of a food source. 2.1 Phases of Artificial Bee Colony Algorithm There are three major steps in search process [11]. These are: The employed bees are sent to the food source and its nectar quality is calculated. On the basis of information from the employed bees about the food source and its nectar quality, food sources are selected by the onlooker bees.. Then determination of scout bees takes place and after that they are employed onto possible food sources. At the initial stage, the bees arbitrarily select the food source location and calculate its nectar quality. The information about the food source with its nectar quality is shared by employed bees with the bees (onlooker) waiting within the hive. After this, employed bees come back to the food sources that have been found in previous cycle. Based on the old food sources, employed bee find new food sources by using its information in the neighbourhood of old food source. At last, the information shared by the employed bees at dance area is used by onlooker bees for finding food source with good nectar quality. The selection of food source and its nectar quality are directly related to each other. The more the nectar quality the more is possibility of selecting the food source. Thus onlooker bees are employed by the employed bees to the food sources having highest value for its nectar quality. Then it chooses the new food source by using the information present in her memory. When the food source becomes abandoned due to the exploitation by onlooker bees, scout bee randomly searches for new food source. [3] Proposed Methodology As discussed in previous section, best food sources are found around the hive because of the foraging behaviour of honey bees. In other words, the flowers with large amount of nectar are chosen by bees. Before explaining the proposed method for robot path planning, first the environment is defined and robot and obstacles is modelled. The working environment is created in such a manner that mobile robot can understand and construe the information necessary for path planning. The working environment is modelled as 2D space. The obstacles are represented as filled rectangles which are of red colour in space as shown in Fig. 1 The first step in making use of bee colony algorithm is to find the initial path for robot to move from source to the destination. The initial path is chosen in such a way that the robot 73

4 Enhanced Artificial Bees Colony Algorithm for Robot Path Planning moves in a direct line from source to the target. Whenever it encounters an obstacle, it moves along the boundary line of the obstacle. Based on position of obstacles in the space, we have breakpoints. Breakpoint is a point that two successive straight lines meet each other. [15] Using this method guarantee that robot will reach the destination but the path obtained is not optimal. So use of bee colony algorithm is done to find the path of shorter length. For this, we have to set the value of various parameters such as number of total bees (n). Out of n, number of employed bees, number of unemployed bees and scout bees are chosen. The number of employed bees is taken to be 75% of the total number of bees while scout and inactive bees are taken to be roughly 15 % and 10% respectively. This constitutes the initialization phase. Subsequently foraging process is initiated. Bees explore and exploit the search space to find the optimal solution for the robot path planning. Each employed bee is put in the neighbourhood of the initial solution. Neighbourhood of the breakpoint is taken at a distance of r which is a random number less than 10. This results in creation of new path. There are two strategies to create new path: a) Replace the previous breakpoint with new breakpoint and now the path has new breakpoint except the replaced one b) Create a path from the source to the new point found and from new point to the destination These two strategies to be performed and path with shorter length should be chosen as new path. Then the fitness of the path is calculated using Euclidian distance. where i,j are two points having x and y coordinates. Now the onlooker bees are sent to the paths found by the employed bees and are put in the neighbourhood of the points found in previous step. Further the bee optimisation is set to find the best solution. Then active bees, scout bees and inactive bees are made to do their part of

5 work. The exploitation process is carried out by employed and onlooker bees while the exploration process is done by scout bees. The active bees and the scout bees observe the waggle dance. The active bee first obtains a neighbour solution relative to its current solution stored in its memory, and then determines the quality of that neighbour. If the current bee finds a better neighbour solution, then algorithm determines if the bee makes a mistake and rejects the better neighbour or if the bee accepts the better neighbour. Similarly, if the current bee did not find a better neighbour solution, the algorithm determines whether the bee makes a mistake and accepts the worse neighbour solution or does not make a mistake and rejects the neighbour. If the bee has exhausted a particular food source without finding the better neighbour solution then active bee is changed to inactive bee. A scout bee generates a random solution, checks if the random solution is better than the current solution in memory, and, if so, copies the random solution into memory. Recall that smaller quality values are better. If the scout bee has found a better solution, the algorithm checks to see if the new solution is a global best solution. An active or scout bee returns to the hive and then performs a waggle dance to inactive bees in order to convey information about the location and quality of a food source. The termination criterion for optimization loop is the number of iterations when completed and the optimal result for path planning is given by the result so obtained. [4] Simulation Results To evaluate the performance of bees algorithm for robot path planning, the above algorithm is used. The environment is shown in Fig 1 with obstacles. Starting point is (0,479) and destination is (479, 0). OpenCv with C++ is used for programming the bees algorithm for path planning. The code has been made to run on 2GHz Core-duo CPU with 3GB RAM. The inputs are environment with obstacles, source and destination position for robot. The initial path found is shown in Fig 2 Fig2 Initial Path The path so obtained is not the shortest path. The length of this path is The value of parameters for bees algorithm are: Number of total bees: 100, Number of active bees: 60, Number of inactive bees: 15, Number of scout bees: 35, Number of visits: 100, Number of iterations: The 75

6 Enhanced Artificial Bees Colony Algorithm for Robot Path Planning results obtained were: average length , maximum length-720.4, and minimum length This path is 26% shorter in length as compared to initial path and contains very less breakpoints. Fig 3 shows the final path of the shortest length. [5] Conclusion and Future Scope In this paper, an enhanced artificial bee s colony algorithm has been presented for solving robot path planning. The result shows that algorithm has shorter length with less breakpoint. Or in other words more smoother path. Neighbourhood search applied after every bee cycle has increased the quality of solution or in other words decreased the distance to be covered during path planning. Tuning of parameters in algorithm will be investigated further. Fig 3 Final path by using bees colony algorithm REFERENCES [1] Lozano-Pe, T., and rez, M.A. An algorithm for planning collision-free paths among polyhedral obstacles, Communications of ACM, 2: , [2] Huang, L. Velocity planning for a mobile robot to track a moving target - a potential field approach, Robotics and Autonomous Systems, 57: 55-63, [3] Siegwart, L. and Nourbakhsh, R., Introduction to Autonomous start Mobile Robots, MIT Press, 1998 [4] Wang, M. and Liu, N.K. Fuzzy logic-based real-time robot goal navigation in unknown environment with dead ends, Robotics and Autonomous Systems, 56: , [5] P. Lucic, and D. Teodorovic, Bee system: Modeling Combinatorial Optimization Transportation Engineering Problems by Swarm Intelligence, Preprints of the TRISTAN IV

7 Triennial Symposium on Transportation Analysis, Sao Miguel, Azores Islands, pp ,2001 [6] Saffari, M. H.; Mahjoob, M.J., "Bee colony algorithm for real-time optimal path planning of mobile robots," Soft Computing, Computing with Words and Perceptions in System Analysis, Decision and Control, ICSCCW Fifth International Conference on, vol., no., pp.1, 4, 2-4 Sept [7] Lebedev D. Neural network model for robot path planning in dynamically changing environment, Modeling and Analysis of Information Systems, 18(1): 12-18, 2001 [8] Noguchi, N. and Terao, H., Multicriteria multistage planning for the optimal path selection using hybrid genetic algorithms, Computers and Electronics in Agriculture, 18: , [9] Guan-Zheng, T., Huan, H. and Aaron, S., Ant Colony System Algorithm for Real-Time Globally Optimal Path Planning of Mobile Robots, Acta automatica sinica, 33: , 2007 [10] Von Frisch K, Bees: Their Vision, Chemical Senses and Language, (Revised edn) Cornell University Press, N.Y., Ithaca, 1976 [11] Seeley T D, The Wisdom of the Hive: The Social Physiology of Honey Bee Colonies, Massachusetts: Harvard University Press, Cambridge, 1996 [12] Bonabeau E, Dorigo M, and Theraulaz G, Swarm Intelligence: from Natural to Artificial Systems, Oxford University Press, New York, 1999 [13]Camazine S, Deneubourg J, Franks N R, Sneyd J, Theraula G and Bonabeau E, Self-Organization in Biological Systems, Princeton: Princeton University Press, [14] D. Karaboga. An idea based on honey bee swarm for numerical optimization. Techn. Rep. TR06, Erciyes Univ. Press, Erciyes, [15] M.H.Saffari and M.J. Mahjoob, Bee Colony Algorithm for Real-Time Optimal Path Planning of Mobile Robots, Fifth International Conference on Soft Computing, Computing with Words and Perceptions in System Analysis, Decision and Control, ICSCCW