An efficient stereovision-based motion compensation method for moving robots is presented in [ 15 ] using the disparity map and three modules: segmentation, feature extraction, and estimation. In the segmentation module, the authors propose the use of extended type-2 fuzzy information theory to recognize the obstacles. Fuzzy logic is used to implement the design and coordination [ 36 ] of a memory grid and to develop a minimum risk method for robotnavigation, and is able to avoid collision with obsta- cles in different scenarios, such as long walls, large concave and recursive UU-shaped regions, unstructured regions, cluttered regions, and maze-like obstacles that repre- sent dynamic indoor environments. In [ 1 ] a fuzzy logic controller is developed based on the Mamdani-type fuzzy method for robotnavigation and obstacle avoidance in a cluttered environment. A fuzzy controller with three inputs and a single output pro- vides safenavigation for the robot motion in a static environment while taking into account the accuracy of the measurements of its position, distance to the obstacles and the goal point, speed, orientation, and the rate of change of its heading angle. The authors in [ 1 ] describe a fast and reliable method of obstacle avoidance for both for outdoor and indoor navigation. The method is applicable in various mobile robotic systems regardless of whic sensors are used and is based on two complementary ap- proaches: non-complex implementation and human-like smooth steering. In [ 37 ] a conceptual approach is considered based on fuzzy logic to solve the local navigation and obstacle avoidance problem for multi-link robots. The fuzzy rule-based approach is considered as an on-line local navigation method for the generation of instantaneous collision-free trajectories.
 H. Kuo, J. Chang and C. Liu, Particle Swarm Optimization For Global Optimization Problems, Journal of Marine Science and Technology, Vol. 14, No. 3, pp:170-181, 2006.  Dian P.R., Siti M.S. and Siti S.Y., Particle Swarm Optimization: Technique, System and Challenges, International Journal of Computer Applications, Vol.14, No.1, pp:19-27, Jan2011.  Shahlla A. AbdAlKader and Omaima N. Ahmad AL-Allaf, Particle Swarm Optimization Based Discrete Cosine Transform for Person Identification by Gait Recognition, The 7th International Conference on Information Technology (ICIT 2015), AlZaytoonah University of Jordan, Amman, Jordan, pp:156- 163, 12 May 2015, doi:10.15849/icit.2015.  Omaima N. Ahmad AL-Allaf, Shahlla A.
We considered the centralized PCE executing different path computation algorithms for the control of single autonomous system. The packets will be segregated in-terms of its class of service category in the LSR. The LSR will send a PCC request to PCE along with the required TE parameters to be optimized. (Ref.Fig.5). Inside the PCE, the QoS metric will get mapped on to the appropriate path computation algorithm. For every path computation algorithm a new child PCE instance will be generated under the control of the parent PCE which orchestrates all the available co-operating child PCE’s. For example, it may be preferred to minimize the economic cost of a path for one LSP while for another LSP the optimization criteria may be minimizing the end-to-end delay.
In this study, Artificial Neural Network (ANN) model has been used for modeling and design of students’ grades in high school level to predict their grades and increase their performance depends on four factors: grades average of ninth level, grades average of tenth level, grades average of eleventh level, and the average of studying hours per day. To do so a Neural Network has been designed. The input parameters were the grades average of ninth stage, grades average of tenth stage, grades average of eleventh stage, and average studying hours per day. One hidden layer was considered with ten neurons. The output layer is Grades average of high school level. One hundred data points were considered to train the network and calculate the weights. After that, the results of ANN model have been used by Simulated Annealing (SA) optimization algorithm to maximize the students’ grades in high school level. MATLAB software was used to do the implementation part. The main goal of this study has been achieved by predicting high school students’ grades, which can help in increasing the students’ performance in this level of education.
In this research, we employ a modularity optimization based approach that is presented by Blondel et al. . The approach consists of two phases that are carried out repeatedly. The first phase finds communities of the current graph, and the second phase builds a new graph whose nodes are the communities found in the first phase and the edges are formed by bundling edges of the current graph in the first phase. The first phase is reapplied to the resulting graph from the second phase and the process continues until there is only one node that contains all the nodes in the original graph. The approach has been shown to be fast and can overcome the resolution limit due to the hierarchical multilevel nature of the algorithm.
technique. This space optimization strategy reduces redundancy problems but do not consider index accessibility problem. Moreover, the backup strategy consumes a huge time overhead over searching hot data blocks. PRESIDIO  strategy uses ARCHIVE data temperature property and reduces data redundancy problem over active part of secondary storage area. Moreover, the approach accesses ARCHIVE section and matches storage objects to reduce space workload. The approach creates a virtual content addressable storage to manage block indexes and generates a huge time overhead to access, copy and paste alike ARCHIVE blocks. Fabien Andre  presents a reliable data temperature storing strategy with erasure coding approach but limits it over homogeneous storage- tier functionality .
Mobile Cloud Computing (MCC) technology enables the resource-constrained Smartphones to outsource its richer applications onto resource-abundant cloud. Cloud computing has emerged as a platform to extend the capabilities of mobile devices regarding energy and resources. MCC offers the services to enhance mobile user convenience, the energy-hungry applications require selective offloading of tasks onto the cloud and optimally scheduling the tasks by taking into account the cloud resources. Even though, the MCC offers the services with high level user convenience, it creates numerous challenges for the service providers. This paper proposes Maximizing User anTicipation on cloUd-based mobile AppLications BesidE NEt proFIT (MUTUAL-BENEFIT), an optimization approach that satisfies the Service Level Agreements (SLA) of the users and simultaneously maximizes the profit of the service provider. The MUTUAL-BENEFIT significantly enhances task offloading, scheduling and resource allocation in MCC environment. Initially, offloading manager of ThinkAir architecture judiciously offloads the intensive tasks based on the knowledge of device energy and performs parallelized execution. Secondly, the non-recursive dynamic programming assisted Ant Colony Optimization (ACO) method schedules the intensive tasks for an extended battery and quick response. It contemplates the SLA objective factors such as execution time, load and profit as the pheromone value of ACO while selecting the optimal resources. Together with SLA- aware task scheduling and the utility function based optimal VM resource allocation maintains an excellent trade-off between the service satisfaction and profit. The experimental results show that MUTUAL- BENEFIT approach preserves the device energy while ensuring the optimal SLA objectives and profit of the provider.
The problem of determining a smoothest and collision-free path with maximum possible speed for a Mobile Robot (R) which is chasing a moving target in an unknowndynamic environment is addressed in this paper. Genetic Network Programming with Reinforcement Learning (GNP-RL) has several important features over other evolutionary algorithms such as combining offline and online learning on the one hand, and combining diversified and intensified search on the other hand. However, it was used in solving the problem of R navigation in static environment only. This paper presents GNP-RL as a first attempt to apply it for R navigation in dynamic environment. The GNP-RL is designed based on an environment representation called Obstacle-Target Correlation (OTC). The combination between features of OTC and that of GNP-RL provides safenavigation (effective obstacle avoidance) in dynamic environment, smooth movement, and reducing the obstacle avoidance latency time. Simulation in dynamic environment is used to evaluate the performance of collision prediction based GNP-RL compared with that of two state-of-the art navigation approaches, namely, Q-learning (QL) and Artificial Potential Field (APF). The simulation results show that the proposed GNP-RL outperforms both QL and APF in terms of smoothness movement and safer navigation. In addition, it outperforms APF in terms of preserving maximum possible speed during obstacle avoidance.
In human societies, trust is dynamic rather than static. Trust grows and destroys based on interactions over time . Currall and Epstein  describe a trust life cycle that consists of three main phases; building trust, maintaining trust and destroying trust. These phases are referred to as evolutionary phases of trust. They propose that in a relationship, building trust starts at a point of neither trust nor distrust due to the limited information on the trustworthiness of a counterpart. This phase builds incrementally as the trustor cautiously observes whether the trust is upheld or violated. Once the trust building behaviours are taken or observed, the trust level continues to grow to maintain the trust phase. In this phase, the trust value is maintained unless a trust-violating event occurs and the move to the ‘destroying trust' phase.
Self-governing analogous parking of a car like mobile robot by a integration of neural and fuzzy based organizer Presented by Demirli et al. . In this paper, they mainly pay attention on toughest container of analogous parking in which the parking planetary proportions cannot recognize. In this the planned representation uses the information from the 3 sonar sensors which is mount on front left place of car to choose steering angle and the 5th-order polynomial position track for generating the training information. Fuzzy representation is recognized by subtractive cluster representation and it is taught by ANFIS and simulating outputs shows that sculpt can effectively choose about the movement way at all variety time with not knowing the parking dimension room width, which is depend on direct readings which is given by sonar sensor that assist as input.
The main application of this work revolves around proposing a real-time and effi- cient way to demonstrate vision-based navigation in UAVs. The proposed approach is developed for a quadrotor UAV which is capable of performing defensive maneuvers in case any obstacles are in its way, while constantly moving towards a user-defined final destination. Stereo depth computation adds a third axis to a two dimensional image coordinate frame. This can be referred to as the depth image space or depth image coordinate frame. The idea of planning in this frame of reference is utilized along with certain precomputed action primitives. The formulation of these action primitives leads to a hybrid control law for feasible trajectory generation. Further, a proof of stability of this system is also presented. The proposed approach keeps in view the fact that while performing fast maneuvers and obstacle avoidance simul- taneously, many of the standard optimization approaches do not work in real-time on-board due to time and resource limitations.
In our first approach to the robotnavigation topic, presented in Chapter 5 , our initial hypothesis is to make the robot navigate according to a human model, the ESFM, and thus, it will be more accepted by humans. We threw the following questions at the introductory Chapter 1 : should robots behave as a person to be considered as equals by people? or should robots take care of humans’ trajectories on their own way? After this work, we conclude that our initial hypothesis is only valid in simple situations (few obstacles and moving people). In more complex situations, the robot is not able to behave like a person, due to the fact that the robotic platform studied in this thesis has wheels, and their propagation model is different than a pedestrian. In addition, we realized that the embodiment of the robot would never be perceived as a person, and hence, it is wrong to make it move as it is a person if none would treat the robot as such. Chapter 5 has motivated a change in the methodology to study robotnavigation from trying to implement a completely characterized human navigation embodied in a robot into implementing a robotnavigation that considers people in the scene developing its own specific strategies to success in social environments.
Moving object detection is one of the most important parts of planning in dynamic environment. The objective is to classify observations as static or dynamic. Researchers developed several methods for this classification. One common method is Expectation Maximization Algorithm (EMA) , . EMA is a two steps maximization process which solves incomplete data optimization problem . Another method is sample-based variant of probabilistic data association filter. This method filters dynamic observations like human and results robust scan matching . Besides, non-probabilistic methods are also developed. For example,  suggested a simple rule for classification. This method is extended for grid map and is used for dynamic observation mapping in this paper.
Reinforcement-learning algorithm in real-time takes a significant long time. Therefore, it is not feasible to run the experiment for too many episodes, even though, a map with small dimen- sions (62cm x 50.5cm) is used (figure 5.12). The number of episodes to run the experiments has been determined by trying the fastest algorithm and checking the time needed for its con- vergence pattern to stabilize. It has been proved that 190 episodes are sufficient to converge to a sub-optimal behaviour for the prioritized smart-planning algorithm. Therefore, other al- gorithms are experimented for 190 episodes to compare them. Making real-time experiments has the problem of non-stationary rewards. Non-stationary rewards are received by the robot due to the skidding. When the robot applies the same action in the same state, it is not possible to move to a unique next-state, but due to the skidding properties, it is probable that the robot will visit a set of next-states. Thus, the reward given by the environment is non-stationary and depends on which next-state the robot will move to. In order to avoid the effect of these noisy rewards, the minimum value of the learning rate α factor is kept smaller than simulations at 0.8. The minimum exploration factor ² is also kept at a large value of 20%. This is particularly useful in case the robot is trapped in local minima. Finally, the action space has been reduced to only "move-forward", "turn-right" and "turn-left" movements. The reason for this is that including the move-backward action increases the probability of getting stuck in local minima during the training. The parameters used for real-time experiments are used in table 5.3.
In this study, we propose a behavior-based navigation architecture, called BBFM, which inherits advantages of fuzzy logic to design the objective functions and multiobjective optimization to fuse the behaviors. In BBFM, each behavior is represented by a reduced fuzzy controller which only contains the fuzzification and fuzzy inference processes. As the result, the output of each fuzzy controller will be a function of input variables whose value presents the achievement of behavior objective, or in other words, the objective function. These functions thus can be used as inputs for a multi objective optimization process to find the optimal control signal. A number of simulations, comparisons, and experiments have been carried out and the results confirmed the efficiency of the proposed architecture in navigating the mobile robot in complex and unknownenvironments.
The A* algorithm can be applied to any graphic representation. However, the path planning becomes more difﬁcult if additional factors such as the topology of the ground is considered. The wavefront algorithm is another example of path planners. It is only suitable for applying to a grid representation of the workspace map [Hughes et al., 1992]. It uses the principle of heating radiation in a conductive material, where the initial position of the mobile robot is considered as a heating source. Thus, if there is any path to the goal, the heat will radiate to it; otherwise obstacles which are considered as having zero conductivity will isolate the heat from reaching the goal. In this method, the free areas are assigned large conductivities, while undesired areas are considered to be low conductivity. To ﬁnd the global path, all obstacle cells are assigned to a ﬁxed value of 1, while the goal is given a value of 2 which is increased to cover all free cells in the world map, and then the path starts from the current position of the robot and follows a sequence of nodes which decreases the numeric value.
The rest of the paper is organized as follows. A review of the related research work in the area of terrain classification is given in section 2.Section 3.describes the proposed methodology for terrain classification for traversability assessment. It explains the feature extraction methods and discusses the classifier used. The details of the proposed path planning algorithm for the robot in navigable terrain are elucidated in section 4.Section 5.presents the experimental results of our traversability assessment of terrain and path planning algorithm and finally section 6.deals with the conclusion and possible extension for the future work.
Most of the future robots will be mobile, and the main challenge is to develop algorithms for their autonomous nav- igation as well as for human-robot interactions. The Laboratory for Autonomous Systems and Mobile Robotics (LAMOR) at the Faculty of Electrical Engineering and Computing of the University of Zagreb is involved in the research of such mobile robotic systems, and currently participates in a number of related international and nation- al research projects. This paper addresses the issue of autonomousnavigation of mobile robots in complex dynamicenvironments, providing state of the art of the domain and major LAMOR’s contribution to it. At the end, we present an application example of the autonomousnavigation technologies in flexible warehouses, which we have been de- veloping within a Horizon 2020 project SafeLog.
In lateral dynamic control of front wheel steer vehicle such as car passenger, the yaw stability control system is one of the prominent approaches to improve the vehicle handling and stability especially during low to mid-range of lateral acceleration. To ensure the lateral stability, it is important to control the vehicle yaw rate where this variable is closely related to the front wheel steer angle. The actual yaw rate response must be tracked and close to yaw rate desired response which generated by a yaw rate reference model. In order to realize this tracking control task, an active front steering (AFS) control can be implemented. The proposed controller will generates the corrective