Edge Servers are servers (computing resources) that do processing at an edge location, which might range from on-premises to regional. it the nearest to the origins of data creation, and it inherits the technologies and functionalities to process data directly at the user's end rather than sending it to a macro data center. the current work studies the effect of the number of sensors on the work of local servers using MATLAB. Then, the proposed system deals with improving the performance of the smart factory. Moreover, focuses on reducing the delay by using hybrid computing, cloud, and edge servers. The factory is represented by a group of sensors (temperature, pressure, gas, level and proximity). that send data to one of the three servers (Two are edge servers with the same copies of data and rules, and the third is the cloud). The sensor values are sent from the factory to the Raspberry Pi 3 and the selection algorithm is implemented to choose only one server that receives the data and to be processed. If two servers are working, the comparison is done by choosing one server, and when one edge server is working, the data is sent to it. But, in the absence of a working edge server, the data is sent to the cloud (web server). These steps are very important to reduce the delay and not to stop the factory when any problem occurs After the data arrives at the edge server, it is processed by making a special decision for each sensitive case in the form of values placed in variables (r1, r2, r3, r4, r5) that are sent to the Arduino, which in turn translates them into making a decision, for example, to turn on or off a specific LED for each sensitive. The time has been reduced using the edge servers. When comparing the total time using one of the Edge servers (2ms) with the total time using the cloud (web server) (128ms), This difference in time seems big and clear as the edge servers succeeded in reaching the smart factory in the least time.

The goal of the free navigating multi-mobile robot is to find the optimal paths to direct their movements, and to avoid the path of other mobile robots, and not to collide with them to reach a specific target, where the two important path planning problems must solve; the first one is that the path must avoid collision with obstacles, and the second one is that it must reduce the length of the path to a minimum. This thesis discusses finding the shortest path with the optimum cost function for the single and multi-mobile robot by using path planning techniques, the Chaotic Particle Swarm Optimization (CPSO) and A-Star, comparesthe results between them and the proposed hybrid algorithm that combines A-Star and CPSO (ACPSO) algorithm. The proposed hybrid algorithm is much better than A-star and CPSO algorithms because it uses the best nodes that are generated from the A-star algorithm and uses these nodes as initial values of the particles in the CPSO algorithm with spline interpolation technique to find the shortest and smoothest desired path for the single and multi-mobile robots. On the other hand, the mobile robot platform has a highly nonlinear kinematics model and time-variant outputs states as well as has an under-actuated system. To improve the actual output trajectory tracking the performance of the multi-mobile robot, a digital convolutional neural network tracking trajectory (CNNTT) controller proposed with off-line and on-line tuning Back-Propagation algorithms to generate precisely and quickly the optimal left and right wheels velocities to track the desired paths equations with a minimum position tracking error and without oscillation response. These algorithms are simulated by the MATLAB package in a fixed obstacles environment to show the effectiveness of the hybrid swarm optimization algorithm in terms of the minimum number of the evaluation function, and the shortest path length. The results of the proposed method showed that the digital (CNNTT) controller is accurate in terms of the tracking ability of the mobile robot to the desired II paths quickly. This performance was achieved by quickly generating smooth linear wheels’ velocity actions for the mobile robot system with a minimum number of cost-function evaluations. This cost function minimized the x-position and the yposition tracking errors to around 4 cm and 2.5 cm, respectively, and minimized the orientation error to approximately zero without oscillation in the responses. Finally, the effectiveness of the numerical simulation results of the proposed control strategy was confirmed through a comparison with other types of controllers’ simulation results.