Cargo handling is a major part of maritime industry. Cargo handlers face risk of accidents from loading and offloading assignment during their work and even worse accidents from the equipment that are used in the shipping industry. Repetitive motion injuries, joint damages and injuries from falls are common accidents particularly for people who move cargo by hand. Furthermore even when cranes are used accidents from mechanical faults or operator errors are also common. In order to reduce accidents in the maritime industry, automation of cargo handling processes is recommended. In particular the use of robots for cargo handling has been tested and implemented with huge success. Autonomous mobile robots have been used to automate some aspect of the shipping industry. A reliable collision avoidance methodology is needed for robots to navigate effectively in seaports. Conventionally robots are fitted with sensors for detecting their environment and transducers for communicating with other robots and the central control system. However the sensing, communication and coordination mechanism are still unreliable for achieving faultless navigation in complex environment like those in seaports. In particular the robot have to overcome pertinent problem of dealing with vagueness in their surroundings.

 

Fuzzy logic is a mathematical method that is appropriate for handling uncertainty emerging from imprecise knowledge. In this work a fusion of mamdani fuzzy inference model which is an advanced version of fuzzy logic is presented. Specifically the model uses eight input ultrasonic sensors and two output variables with twenty-seven fuzzy rules to solve the navigation problem. The study investigates possibility of modelling and handling uncertainty by tuning controller and applying sensor data to achieve efficient results for mobile crane navigation in seaport. Multiple simulations were performed to validate and to check the feasibility and efficacy of the proposed model. The implementation was performed in V-REP and MATLAB software. The results of this work provide a promising solution that will find application in the marine industry for automation of cargo handling. Large scale automation will improve the efficiency of the seaport and contribute to an increase in the volume of trade thereby leading to economic growth.

 

Introduction.

Autonomous robot navigation constitutes one of the major trends in robotics research. This is inspired by existing gap between available technologies and different application demands. Existing industrial robots have low flexibility and autonomy. Normally, these robots implement pre-programmed sequence of procedures in highly controlled environments, and are not able to function in new situations or face unpredicted states. Table 1 shows several algorithms suggested highlighting pro and cons.

Table .1 Comparisons of obstacle avoidance approaches.

  Approach Advantages Limitations
1 Potential fields

 

 

 

 

 

 

[1]  [2]

-superior in contrast to global path planning

-Simplicity and mathematical elegance.

-When attractive force and repulsive is equal or almost equal but on opposite direction the potential force of the robot is zero thus trapped in local minima or oscillation.

-When robot is far attractive force become very great leading robot to move close to obstacles. Therefore  risk of collision to obstacles

-Goals are non-reachable with obstacle nearby (GNRON)

-Do not perform well in complex scenario having many obstacles.

2 Neural networks

 

 

[3] [4],

-Minimization of time required for training the network

-Ability to learn and model non-linear and complex relationships

-After learning from the initial inputs and their relationships, it can infer unseen relationships

-No clue how results are generated, black box, so if you want to know what causes the output you can’t with a neural network.

-Need a larger data

–Hardware requirements is large

-Suitable for complex problems else costs overweighs benefit

3 Genetic algorithm

 

 

 

 

 

 

[5]

-Easy to understand

-Solution gets better with time.

-Optimization good for noisy environment.

 

-No guarantee of finding global maxima but likelihood of getting stuck in a local maxima

-Need sized population and a lot of generations before you get good results.

-Fine tuning all the parameters for the GA, like mutation rate, elitism percentage, crossover parameters, fitness normalization/selection parameters, etc., is often trial and error.

-communication to the system is through the fitness function. The result could be crazy, inefficient or incomprehensible

4 Vision based navigation

 

[6]or explicit knowledge of the structure of the crop rows. This method works by extracting and tracking the direction and lateral offset of the dominant parallel texture in a simulated overhead view of the scene and hence abstracts away crop-specific details such as colour, spacing and periodicity. The results demonstrate that the method is able to track crop rows across fields with extremely varied appearance during day and night. We demonstrate this method can autonomously guide a robot along the crop rows.”, “author” : [ { “dropping-particle” : “”, “family” : “English”, “given” : “Andrew”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Ross”, “given” : “Patrick”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Ball”, “given” : “David”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Corke”, “given” : “Peter”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } ], “container-title” : “Proceedings – IEEE International Conference on Robotics and Automation”, “id” : “ITEM-1”, “issued” : { “date-parts” : [ [ “2014” ] ] }, “page” : “1693-1698”, “title” : “Vision based guidance for robot navigation in agriculture”, “type” : “article-journal” }, “uris” : [ “//www.mendeley.com/documents/?uuid=150e7bb7-f75c-43a6-964e-5933c50d3e54” ] } ], “mendeley” : { “formattedCitation” : “[6]”, “plainTextFormattedCitation” : “[6]”, “previouslyFormattedCitation” : “[7]” }, “properties” : {  }, “schema” : “//github.com/citation-style-language/schema/raw/master/csl-citation.json” }. [7].

-Accurate

-Less noisy.

-locating obstacle fail when the mark cannot be seen

-The main disadvantages of this approach is in terms of cost and challenges related to installation

5 Fuzzy logic

 

 

 

 

 

[8]. [9]

-Easy to model the reasoning.

-The ability to deal with uncertainty and nonlinearity.

-Ease of implementation and use of linguistic variables.

-Mimics human control logic.

-Use imprecise language and are inherently stable.

-Flexible and can be modified easily.

-Can be combined easily with conventional control techniques.

 

-logic requires experimentation and experience.

-Finding the most appropriate function can be found by trial and this can take quite a while.

 

 

A clear market is emerging for strictly autonomous robots. However, the biggest challenge is path finding and motion control. Fuzzy logic has demonstrated to be an appropriate tool for handling uncertainty and knowledge representation [10]and in recent years the use of mobile robots in material handling has considerably increased. Usually workers push carts around warehouses and manually handle orders which is not very cost-effective. To this end, a potential method to control a swarm of mobile robots in a warehouse with static and dynamic obstacles is to use the wireless control approach. Further, to be able to control different types of mobile robots in the warehouse, the fuzzy logic control approach has been chosen. Therefore, in this paper, an on-line navigation technique for a wheeled mobile robot (WMR [9][11]. Fuzzy logic method for obstacle avoidance was presented [12],robot reacted to environment accordingly with the aid of sensors. However in some occurrences sonar sensors failed to detect certain obstacles. Research based on ultrasonic sensors demonstrated failure in detection of objects with cylindrical and spherical shapes. The sonar beam hit the surface with oblique incidence and reflects away instead of going back as an echo thus no detection   [13]  [8]. This study increases scanning angle by fusing eight sensors for wider detection range. A review of several investigations carried out on the use of fuzzy controllers’ revealed numerous approaches previously developed for path planning and control of obstacle avoidance mobile robots [14]. Fuzzy logic emerges suitable for mobile robot because of robustness and ability to handle uncertainties. This study investigates possibility to maintain uncertainties using various fuzzy logic membership functions and application of wider detection angle. Simulations are performed using VREP and MATLAB software to validate effectiveness of the projected control model.

 

Theoretical concept.

In real navigational process, mobile robot need to match up obstacle evading action and path tracking control. Realization of autonomous obstacle avoidance is anticipated if the mobile robot make suitable responses and attain a collision free path based on surrounding information perceived by sensors. Fuzzy controller mimicking human being driving intelligence is considered in this study. The conceptual framework is illustrated in Figure 1(a). In the process of mobile robot avoiding obstacles, fuzzy controller generates control command. Fuzzy controller adjusts robot movement to a collision free path. Depending on information acquired by sensors relevant fuzzy control rules are activated. The outputs of activated rules are combined by fuzzy logic operations to control velocities and steering angle of robot wheels. Flowchart of navigation process is shown in figure 1(b). A total 27 fuzzy rules were built as summarized as in Table 2. The general rules fulfil the following points:

When there are no obstacles in the surroundings or obstacle are far away, left wheel and the right wheels  moves faster, i.e., the output is “BIG”;

When the mobile robot detects obstacles, appropriate reaction is made by changing velocities of driving wheels. Direction changing action is determined according to the obstacle distance and obstacle orientation relative to the robot.

 

The inputs of fuzzy obstacle avoidance controller are responsive sensing distances dl, df, and dr.controller determines the action of mobile robot according to sensing distance. The outputs of the fuzzy obstacle avoidance controller are velocities vl, vr for the two driven wheels. The inputs are taken as premise variables with three fuzzy linguistic sets labeled as “NEAR (N)”, “MEDIUM (M)”, and “FAR (F)” with discussion region of  [0, 1000mm]. The output variables are represented by three fuzzy linguistic sets labeled by “SMALL (S)”, “MEDIUM (M)”, and “BIG (B)”  with discussion region of (0 to 100 cm/s).

 

Design

Eight transducers are used each scanning 230 angle thus to cover approximately 1800 scan for obstacles as shown in Figure 2. When an obstacle existed in the detectable range of any sensor, it measures the nearest distance to the obstacle. Then, the distance between mobile robot and an obstacle is determined by coordinating all sensing feedback data. In the first model three ultrasonic sensors are mounted at points A B and C respectively. The next model has eight transducers. Results presented fused model as most effective [15].

 

Discussion

To monitor the Effects of changing input and output membership function of the model several variables in controller design are adjusted. When the peak value of a membership function are changed used rules changed the fuzzy label. Changing the width value of a membership function affected interpolation between the peak value of function and its adjacent membership function. Experiments were conducted to investigate the membership function shape effects on the mobile robotic performance [16]. The experiments involved changing the shape of the input and the output membership function of Fuzzy controller. Five trials were conducted for each combination of input and output membership function and arithmetic mean used for validation based on time used to reach the destination as shown in table 3.

 

Closer inspection of table 4.4 showed the best performing model was triangular input MF with trapezoidal output MF in average environment. The longest time to reach the target was experienced in simple environment using triangular MF and Gaussian output MF. The most striking observation that emerged from the data comparison was that time to reach the target was shortest in average environment which was rather unexpected results. Surprisingly it took longest time in simple environment with only four obstacles. Gaussian input membership function performance displayed in figure 4.12 shows robot took shortest time to reach the target in average environment this was accomplished using trapezoidal output MF. No significant differences were found between simple and complex environments using trapezoidal output MF. Results of this study shows that triangular input MF and trapezoidal [9]–[11]and in recent years the use of mobile robots in material handling has considerably increased. Usually workers push carts around warehouses and manually handle orders which is not very cost-effective. To this end, a potential method to control a swarm of mobile robots in a warehouse with static and dynamic obstacles is to use the wireless control approach. Further, to be able to control different types of mobile robots in the warehouse, the fuzzy logic control approach has been chosen. Therefore, in this paper, an on-line navigation technique for a wheeled mobile robot (WMR was based. In [9]  efforts to reduce uncertainties concentrated on effects caused by inaccuracy of sensors. In our study similar component features but a lot of attention was on fusion and revision of membership function. Experimental studies conducted by simulation using different  computer’s having dissimilar speed and processing power found that time taken for robot to reach the target differed . Further work is required to establish the viability of involving real robotic crane for both screening and monitoring in an industrial application. Ultrasonic sensors use sound, they are completely nonfunctional in a vacuum hence infrared sensors are suggested for prospect research. This study concentrated on stationary environment future work should consider dynamic environment with moving obstacles. The robot model moved in forward direction only that made robot stack when surrounded by obstacle. Further research should be undertaken to investigate a robot that can also move in reverse direction to eliminate local minima.

 

Conclusion

This work has successfully demonstrated effect of various membership functions in the fuzzy control of mobile robot, different types of membership functions were tested. Consideration was taken for the most three popular types which are the triangular MF, trapezoidal MF & Gaussian MF used as both  input and output variables. The response was analyzed and compared and it shows that the triangular and trapezoidal MF are good in response in terms of rise time taken to reach the target. While the Gaussian MF was poor in response in all cases and this prove the theory that Gaussian MF are better in systems dealing in data on probabilities and statistics. Future work involve testing using different sensor and implementation of the hardware.

Leave a Reply