fuzzy logic, neural networks and evolutionary computation paradigms have confirmed to be effective and potential for smart control of composite systems. Specifically, fuzzy logic has demonstrated to be an appropriate tool for handling uncertainty and knowledge representation [1, 2,3]. A clear market is emerging for strictly autonomous robots. Potential applications include service robots that are intelligent for offices, factory floors and hospitals; robots operating in hazardous or areas that cannot be accessed easily; domestic robots for housework or entertainment and semi-autonomous vehicles to support disabled people in the society. However, the biggest challenge is path finding and motion control. Path planning can be categorized into global and local planning methods . In global planning technique, the mobilerobot prerequisite is that environment should be entirely known and stationary. Contrary local path planning robot need the surroundings to be partly known or totally unknown. The autonomous robot make use of received sensory information in the course of its local navigation . Set of actions are activated to make wanted performance . Robot navigation methods can also be classified as model based method [7,8],fuzzy logic based method [9,10, 1] and reactive method based on neural networks [11, 12].
In this section we compare and evaluate the performance of both Mamdani and Sugeno FLC in term of the path travelled, the smoothness and the efficiency of the FLC for single and multi-robot. For integration of FLS for obstacle avoidance behavior of mobilerobot, the simulation begins upon the successful remote application program interface (API configuration) between V-REP and MATLAB. If the distance from the target and robot position is more than 0.1m, the robot is far from the target location. So, the robot will start move towards the target with three conditions. The second condition is, if the angle is bigger than 0.1 radian the robot will rotate left whereas the third condition is if the angle is smaller than 0.1 radian the robot will rotate right. While executing these three conditions, if any of the proximity sensors of roots detects the obstacle, the FLC either Mamdani or Sugeno, will control the left velocity and right velocity of robot based on the 15 rules defined. This process will keep on repeating until the distance between the robot and target location is not more than 0.1m.
Moreover, in any neural network navigation algorithm, if the robot platform or the type or number of the sensors are changed or altered, the network architecture requires some modifications to accommodate with the new amount of sensor data. In other words, same network architecture cannot be used for different robot platforms with dissimilar types or different numbers of range sensors. Therefore, these network structures will only be effective for the mobile robots that they have been designed for and are not extendable to other sensor configurations. For example, if a neural network is designed to have eight inputs from eight ultrasonic sensors, then this network is not operational for a robot which has only four ultrasonic sensors. Same situation occurs with different types of sensors. For instance, a network which is designed to function with one type of sensor (e.g. a laser scanner) cannot be applied to robots with other types of sensors (e.g. ultrasonic sensors). As a result, the structure of the network needs to be changed and new samples need to be gathered in order to accommodate with the new configurations.
Let us consider measuring the feet position with a 2D range sensor to estimate the walking pattern using as little information as possible. Because the range sensor can only measure the shortest distance to the objects, both legs cannot be measured at the same time, i.e., the idling leg motion in Stage 1 and Stage 3 cannot be measured. Changes in the measurement is plotted as a square-wave along the direction of movement from side to side. This method is thus not directly appropriate for use in the above walking model. However, to execute col- lision avoidance control, real-time velocity data for the idling leg are required. Thus, a Kalman Filter is applied to estimate the unmeasurable idling leg motion.
This chapter covers on the applications of mobilerobot and the summary and discussion of section. On the applications of mobilerobot, it covers briefly on obstacle avoidance in indoor environment. Next, the type of distance sensors, the type motors, and the type of controllers are covered in obstacle avoidance in indoor environment. The different types of distance sensor and motor used for obstacle avoidance and their construction are also include in the type of distance sensor and type of motor sections. Other than that, different types of controller are reviewed in order to control the mobilerobot to avoid the obstacle.
Abstract: The paper is discuss on develop and implement a vision based obstacle avoidance for mobilerobot using optical flow process. There are four stages in this project which are image pre-processing, optical flow process, filtering, object stance measuring and obstacle avoidance. The optical flow process are an image resizing, set parameters, convert color to grayscale, Horn-Schunk method and change grayscale image to binary number. Next process is a filtering done by smoothing filter then image center will be defined. The maximum distance object from a camera has been set as 20 cm. Therefore, the decisions of the robot to move whether left or right are based on the direction of optical flow. This avoidance algorithm allows the mobilerobot to avoid the obstacles which are in different shape either square or rectangular. A friendly graphical user interface (GUI) had been used to monitor the activity of mobilerobot during run the systems.
There are two type of FLC, Takagi-Sugeno  and Mamdani  controller. The efficiency of these two controllers for obstacle avoidance behaviour of robot has been studied by many researchers. However, there are less contribution towards the comparison between these two controllers for obstacle avoidance behaviour of mobilerobot. So, comparative comparison between Mamdani and Sugeno type FLCs for obstacle avoidance behaviour of mobilerobot able to contribute for future works.
Fig. 8 shows the ROS graph for communication between four nodes. The topic distance_detect is initialized by sensor node and subscribed by servo motor node to detect the obstacle distance. Besides, the servo motor initializing three topics call dist_detect_right, dist_detect_left, and dist_detect. The topics of dist_detect_right and dist_detect_left are subscribed by compare distance node to compare the distance of left and right, and then publish a message to DC motor by through direction topic. Whereas, the dist_detect topic is subscribed by DC motor node to stop the robot from moving. The DC motor node also initializes a topic called forward to publish a move message to servo motor node. B. Execution of ROS Nodes
We used an RF-transmitter to be able to send commands to move the robot. Material used to prototype electronic components that would fit onto the Arduino Uno board using the header rails as mounts. A small resistor divider network and a hand full of pushbuttons were utilized to build a circuit that was read by the analog- to-digital converter (ADC) of the Arduino. These ADC values where translated into movement commands and sent to the robot over an RF communication channel to make the robot move in different directions such as forward, backward, etc .
The main objective of this study is to design and construct an autonomous robot. It is two identical mobile robots. It means that the robots are communicating to each other and will collaborate to accomplish specific task given. Each robot is equipped with infrared sensors to detect and avoid obstacles around the environment when perform the task following.
There has been a great effort in studying mobile robots and their locomotion system. In literature, robots are usually classified according to the type of its locomotion as: wheeled, tracked, legged and hybrid. As reported by Rodriguez et al. , a number of developed prototypes tried to mimic the walking in nature which is a complex task involving mechanics, electronics, sensing and control. Wheeled and tracked mobile robots can perform fast and robust motion on flat terrain, but are ineffective to move over uneven terrain or low friction, such as sand and mud.
The field of autonomous mobile robotics has recently gained the interests of many researchers. Due to the specific needs required by various applications of mobilerobot systems (especially in navigation), designing a real-time obstacle avoidance and path following robot system has become the backbone of controlling robots in unknown environments. Therefore, an efficient collision avoidance and path following methodology is needed to develop an intelligent and effective autonomous mobilerobot system. Mobile robots are equipped with various types of sensors (such as GPS, camera, infrared and ultrasonic sensors); these sensors are used to observe the surrounding environment. However, these sensors sometimes fail and have inaccurate readings. Therefore, the integration of sensor fusion will help to solve this dilemma and enhance the overall performance.
In addition, it was found that most sources ((Wang, Xu, Guzman, Jarvis, Goh & Chan 2001), (Sahin & Gaudiano 1998), etc.). Were irritatingly glib concerning the actual method used to avoid obstacles. Most of the sources devote the majority of their time to extracting the relevant information, not discussing exactly how it would be used. Thus very little information could be obtained on the best method of using the Looming data. In the event this was not critical, as the Looming results were too unstable to be employed anyway. However for the sake of completeness there follows a brief description of the planed avoidance method. This was never developed fully as it became clear from a relatively early stage that it would not be implemented.
B. Obstacle Detection and Avoidance using SONAR sensors A SONAR, Sound Navigation and Ranging, unit consists of an ultrasonic oscillator and a receiver. The oscillator pro- duces sound waves at a particular frequency. When these sound waves encounter obstacles, they reﬂect back as echos that are received by the receiver. The output of the sonar receiver is a distorted wave having a 10-20 mV peak to peak voltage. The received voltage varies proportionally to the strength of the signal. By calculating the time taken by the sound waves to reach the receiver after they have been transmitted, the distance of the object from the transmitter can be calculated accurately .
In this paper, we investigate a means by which a machine vision system could utilise optical blur as an avoidance indicator. The methods used are intended for monocular systems and employ the blur recovery methods of Hu and de Haan to find optical blur and optical blur and the looming method described by Raviv and Joarder and Sahin and Gaudiano to relate this to object approach. It was intended that this system be relatively simple in hardware and software implementation. To verify the success of the design, we conducted tests in a controlled environment. It was found that obstacle approach could reliably be computed through this method, but its success depended on the camera lens properties.
The aim of this project is to develop an avoidance behaviors program for a mobilerobot that consists of 4 legs that employs 8 servo motors. A PIC Microcontroller is been implemented to act as the brain for the robot that controls the walking and turning algorithm. The system and the programming will be able to control the movement of the legged so the robot able to walk straight ahead, make left turns and avoid obstacle.
 张凤，谈大龙．一种基于相对坐标系下移动机器人动态实 时避碰的新方法 [J]．机器人，2003，25(1)：31-34,79. Zhang Feng, Tan Dalong. A new real-time and dynamic colli- sion avoidance method of mobile robots based on relative coor- dinates[J]. Robot, 2003, 25(1)：31-34,79.
Mobilerobot design is about art and individual skill to create the useful robot for human application. Each part of mechanical, electrical and software should be studies to make sure that the alls mobilerobot application can run smoothly and can complete the task given. In mechanical part, each measurement of the mobilerobot design must be details and fixed to make sure that the cost is not changes and affect to the whole process design. In electrical and software part also need research because the price is quite expensive and more sensitive if compare with the mechanical. Each equipment specification must be study details from it design, application and until the prices . Figure 2.1 shows the example of mobilerobot design with the range covered by sensor.