Model-based AGVs Scheduling for Assembly by LMS. This scheduling solution focused on addressing the AGV scheduling problem using a model- based approach. In specific the AGV schedule could be fully represented in a mathematical way using MIP and Fuzzy logic for the description of AGV positions along the factory grid. At the first development phase, Pyomo library was used in order to formulate the problem in Python. However, the proposed model was a non-linear MIP limiting the solvers that could address this problem into a small portion of all the solvers available on the market. This was found not to be a valuable solution and thus Genetic Algorithm was used instead in order to approximate the problem’s solution. PyGAD library was used with built-in functions for implementing the GA in any kind of optimization functions. It is important to mention that unlike constraint- programming solutions, metaheuristics do not have a build-in capability for constrains. In order to address this issue, the objective function was altered in order to give bad results (penalties) when the problem constrains are broken. The following figure shows the optimization process of the algorithm for a simple example, reaching 92.22% in 35 iterations for a population of 80 genes. Figure 17 Optimization process of the GA represented as fitness versus genetations; the negative fitness is caused due to the penalty applied for breaking the problem constrains Before GA, PSO algorithm was used for the same problem however did not achieve the same efficiency; despite the good efficiency of GA, it stills provides some latency in the results due to the non-linearity of the model. Additional efforts will be put during the deployment phase of the project in order to improve this model and provide lower response time and accuracy to this problem. The output results were provided in the form of a binary decision variable including several transportations that the AGV performed at a specific point in time (shown in figure below).
