Robot Global Path Planning Based on Improved Artificial Fish-swarm Algorithm

In This study, a new artificial fish-swarm optimization, to improve the foraging behavior of artificial fish swarm algorithm is closer to reality in order to let the fish foraging behavior, increase a look at the link (search) ambient, after examining environment, artificial fish can get more status information of the surrounding environment. Artificial fish screened from the information obtained optimal state for the best direction of movement. Will improve the foraging behavior of artificial fish-swarm algorithm applied to robot global path planning, including the robot to bypass the analog obstacles selected three ways: go obstructions outside, go inside the obstacles, both away obstructions and went outside obstacles Thing achieve robot shortest path planning. Via the MATLAB software emulation test: the improved foraging behavior of artificial fish-swarm algorithm to improve the rapid convergence of the algorithm and stability, improve fish swarm algorithm to the adaptability of the robot global path planning.


INTRODUCTION
Robot has used in many-sided at present such as industry, agriculture, service and aerospace, etc.All of these can be seen that the implement of robot.It needs to be moved to complete specific task in certain profession.So the robot path planning has become a hot issue.According to the partition of current academic circles, robot path planning can be divided into two types: global path planning based on environment information completely known and part path planning based on sensor information.However, the environment modeling of global path planning include: Visible Graph, Topological Method, Free Space, neural network method, etc. Free Space possess advantages, such as model simple, good achieve result.But it is difficult to model for Free Space in a complex environment.Some scholars have researched the robot global path planning based on artificial fish-swarm algorithm, such as the based on artificial fish-swarm algorithm.The study is provided a way which the artificial fish-swarm optimizes robot path planning to Free Space (Nie and Zhou, 2008).Particle swarm optimization work in the aspect of robot global path planning is also to be researched by some scholars, such as the based on the mobile robot path planning of particle swarm algorithm.The study offered a particle swarm algorithm path planning based on Free Space (Huang et al., 2006;Wei, 2012).However, these two papers are searched with low precision, inefficient and other disadvantages.In order to improve the precision and efficient, it has to be the necessary improvement.
The application approximation algorithm to the energy consumption of the robot, considering the energy consumption of the linear motion, but did not consider the energy at the time the, the robot speed change and cornering consumption (Brooks and Kaupp, 2007;Zheng and John, 2006).The Optimization through genetic algorithms to optimize the robot's movement speed and energy consumption (Brooks, 1986;Wang et al., 2009).Using EGICA (Environment-Gene evolutionary Immune Clonal Algorithm) algorithm for robot joint trajectory planning, get to meet the optimal time and optimal energy trajectory path length constraint is not embodied (Xu et al., 2010).The assumption that the robot along the linear motion and speed does not change, is proposed an energy-saving motion control strategy to minimize the consumption function, due to the engineering robot is not along a linear motion, so is difficult to apply to the actual (Barili et al., 1995;Vasilyev, 2002).Order rate control method, using the proposed iterative search algorithm the saving speed change map, reducing the energy consumption of the robot (Chong and Byung, 2005;Xie et al., 2007).It can be seen, the current study is based on a point or two constraints path planning and rarely consider turning and speed changes.Therefore, in accordance with the actual situation of the robot traveling, planning a meet time, path and energy consumption of three the relative optimal path constraints, both in terms of energyefficient or real-time algorithm has great significance.
Artificial fish-swarm is a swarm intelligence algorithm and the late-model intelligence algorithm.
Simplicity, global superiority, rapidity, shadowing property and robustness.The individual of artificial fish hunt object by bunching behavior, rear-end behavior, foraging behavior and random behavior (Li and Qian, 2003).According to these four behaviors, artificial fishswarm can transfer fish-swarm unconsciously to achieve more high-efficiency swarm search.But the fish-swarm algorithm is also has disadvantages with low convergence precision and low hunt efficient.Solve the maximum problem of function: Maximum value point of this function is (0, 0).How to work out maximum value with fish-swarm algorithm.There is a obvious disadvantage in traditional fish-swarm algorithm that maximum value convergence precision is low.The fish-swarm algorithm of improving foraging behavior has resolved it.Improved foraging behavior pseudo-code is following: In Fig. 1, blue point expressed artificial fish.It can be covered that only an artificial fish arrive at maximum point.The data of call-board display as followed is simulation result: maximum point (0.046247, 0.005745), maximum value f(x, y) max = 0.9996.
The parameters define of artificial fish improved foraging behavior as followed: foraging behavior random search k = 20, the visual scope visual = 3, step length step = 0.3, crowding factor δ = 0.618, iterations is 50, 10 artificial fishes.Mapping result with MATLAB 6.5 is followed Fig. 2.
It declared that most artificial fish are rid of what sub maximum focus around maximum.The result of call-board is (-0.023021,0.007922), f (x, y) max = 0.9999.The improved foraging behavior artificial fish has optimal research result and high precision.There are more artificial fish get to maximum point.
From simulate result, it can be concluded that: if foraging behavior search surrounding environment randomly only one time, artificial fish implement foraging behavior or not is provided with contingency.

THE RESEARCH OF ROBOT SHORTEST DISTANCE WORK PATH
System describe: In the Fig. 3, robot obstacle avoidance coordinate, red circle is the destination of robot needs to arrive.It needs to walk 5 places to hunt for the shortest path.Thus, how does it search the shortest path?We resolve the way of robot pass the obstacle.There are three ways for robot to pass obstacle:  There is only a side exist obstacle on robot walk path.That is walk top left corner when A-B  There are obstacles on both side of walk path  The walk path is consists of both A and B, that is to say some sections are first type path, others are second type path It need to confirm the fish if artificial fish is used to hunt for the shortest path.Point P 1 and P 2 express fishes, that is X =(P1, P2)= (x1, y1, x2, y2).L expresses fish food .
The distance d between two fishes is: The fish centre X C is:

      
There is no enough time to wait myself code fishswarm algorithm to hunt for the shortest path.So the description is used mathematics.

Structure second artificial fish-swarm model:
There is another choice is walk inner line (obstacle exist both two sides).It must through segment S 1 T 1 and segment S 2 T 2 .So the path of A to B is: connection of point A and q1 which on segment S1T1, connection q1 and q2 which on segment S2T2, connection q2 and B, these are three segments are consist of route of A to B. S1, T1, S2, T2 are known point.Use q1, q2 to express them: 1, 2 ( 0 ,1) t t  Mathematic description: Artificial fish X = (q1, q2), the quantity of fish  , The distance d between two fishes is: The fish centre X C is: Artificial fish, the quantity of food, fish centre have confirmed.The shortest path can be worked out if reasonable step, visual scope and crowding factor δ are ruled.
The third type robot is consists of first type and second type.So it can work out the shortest path by the solution of first type and second type if divide third type robot into first type and second type.It is stated no longer.

THE ACHIEVE OF SIMULATION
Program run environment is AMD Sulong dikaryon 2.91 GHz, OS is Windows XP, program executive software is Microsoft Visual C++ 6.0.
The shortest path of type a method: At first, we supposed type A is segment from initial position (1, 7) to destination 1(5, 12.5).It can discuss with two ways.
One is walk outboard (outer line); other is inner line which the components are object M, object N and object H.It will through two points: P1 and P2.P1 and P2 need artificial fish to confirm the most optimal position.
The parameter of artificial fish is: artificial fish m = 50, visual scope Visual = 0.15, the largest step length step = 0.5, crowding factor δ = 9, iterations is 100, the initial value of artificial fish is any value in reasonable range.
Operation artificial fish-swarm with MATLAB map can get the result as Fig. 4.
The shortest path of type B method: Supposed that type B is segment from initial position (1, 7) to destination 2(9.8, 14).It can discuss with three ways.
One is walk outer line of object M (path 1, shown as path 1 of Fig. 5).One is third type path which consists of object M, object N, object H (path 2, shown as path 2 of Fig. 5).The last is shown as path 3 of Fig. 5.The path 1 of artificial fish is X1 = (A1, A2), food is the total length of path.The path 2 of artificial fish is X2 = (C1, C2), food is the total length of path.Artificial fish distance d and centre XC are formulation fit with 3.2 th chapter.It is stated no longer.The parameter of artificial fish is: artificial fish m = 50, visual scope Visual = 0.15, the largest step length step = 0.5, crowding factor δ = 9, iterations is 100, the initial value of artificial fish is any value in reasonable range.Mapping result with MATLAB is shown as Fig. 6.

CONCLUSION
Although come up with artificial fish-swarm very early, previous researches are localized.It does just optimize the algorithm main to gain the precise astringency path.Most artificial fish can near around the optimal value.The astringency of robot walk path is more precise.So we choose improvement fish-swarm algorithm that is increase the search surrounding environment times of foraging behavior.It is gets rather improvement of fish-swarm algorithm precision to foraging behavior.But there is little difference on as astringency precise.On the optimizing of robot path, there isn't obvious difference in these two algorithms as the issue is rather simple.However, there is evident distinction on these two astringency precisions of search function s' maximum problem.Improved foraging behavior fish-swarm algorithm astringency is more precise.

ACKNOWLEDGMENT
The authors wish to thank the helpful comments and suggestions from my teachers.This study is supported by the authors are highly thankful for the financial support of National Defense Basic Research Project (B3720110008) and Guangxi Autonomous Region Education Department General Project (200103YB137) and Hechi College of key projects (2010YBZ-N001).

Float
Artificial-fish::AF-prey () { Xmax = Xi; Ymax = Yi; // Xmax Ymax express fish and food quantity which searched in visual scope and with the maximum food for (i = 0; i<try number; i++) //try_number is the movement of artificial fish.{ for (j = 0; j<20; j++) //Hunt for 20 visual scope states.{ Xj = Xi + R and ()*visual; //Search visual scope states randomly.If (Ymax<Yj) //The hunt state possess food quality more than current food quality Ymax.{ Xmax = Xj; //It satisfied condition and implemented what the state Xj assigned to Xmax.Ymax = Yj; //It satisfied condition and implemented what the state Yj assigned to Ymax.} } if (Ymax>Yi) // To judge the searched optimal state value is bigger than former food state.Xi|next = Xi + R and () *step* (Xmax-Xi) /||Xmax-Xi||; //Move to the most optimal direction.else Xi|next = Xi + R and () *step; // Or it move randomly.} return AF-foodconsistence (Xi|next); // Return the state value after fish finish its movement.} The operating environment of following code is: program run environment is AMD Sulong dikaryon 2.91 Ghz, OS is Windows XP, program executive software is Microsoft Visual C++ 6.0.It just searches randomly one time in tradition fish-swarm algorithm foraging behavior.The visual scope visual = 3, step length step = 0.3, crowding factor δ = 0.618, iterations is 50, 10 artificial fishes.Mapping result with MATLAB 6.5 is followed Fig. 1.

INTRODUCE RELEVANT DEFINITION Relevant definition: Definition 1: Vector
x i represents the i th artificial fish state.The x inext represents the i th artificial fish percept next state.In perception range, artificial fish i searches next state x inext random.Artificial fish update state if the food concentration Y inext is superior to the food concentration of current state Y i , or it implements random behavior. Bunching behavior: The current state of artificial fish i is x i and the range of artificial fish i percepts the home position food concentration is Y c .Calculate the quantity of artificial fish n c of perception scope.If formula Y c /n c >δY i and Y i <Y in current perception range and search the number of artificial fish n f in perception range.If Y j /n f >δY i and Y i -Y j then it shows there are more food and not too crowd.Artificial fish i forward into artificial fish j a step, or implement the random behavior. Random behavior: Achieve random behavior is simple.Choosing a state randomly in visual field, move a step to this direction.It is a simulation behavior to natural fish.It ensures that the artificial fish flee sub maximum to hunt for maximum point.
1 , x 1 ) = || x 1 -x 2 ||.Definition 7: Definite a variable to record the optimal history what artificial fish searched.The variable is just like a call-board.So, it is called call-board variable.Defined the artificial fish relevant behavior:  Foraging behavior: c established, it indicates there are lots of food and without crowd.Then artificial fish i forward into home position a step, or implement the random behavior. Rear-end behavior: Artificial fish i current state is x i , search the most concentrative individual artificial fish j Improvement foraging behavior: It just can search around randomly one time for previous artificial fishswarm algorithm.It can't usable reflection surrounding environment conditions to this once random research.