Kudret Demirli

Subtractive clustering based modeling of job sequencing with parametric search

In this paper, an extended subtractive clustering based fuzzy system identification method and the Sugeno type reasoning mechanism are used for modeling job sequencing problems. This approach can be used to build a fuzzy model of the sequencing system from an existing sequence (output data) and possible job attributes (input data). The single machine weighted flowtime problem is used as an example to demonstrate the proposed methodology. The effects of data scarcity on the modeling performance is studied by using three data sets with Varying degrees of available data. Furthermore, a parametric search on various clustering parameters is performed to identify the best model. As a result of parametric search, ranges of clustering parameters that provide best models are also identified.

A genetic approach to two-phase optimization of dynamic supply chain scheduling

In todays competitive environment, agility and leanness have become two crucial strategic concerns for many manufacturing firms in their efforts to broaden market share. Recently, the build-to-order (BTO) manufacturing strategy is becoming a popular operation strategy to achieve both in a mass-scale customization process. BTO system combines the characteristics of make-to-order strategy with a forecast driven make-to-stock strategy. As a means to improve customer responsiveness, customized products are assembled according to specific orders while standard components are pre-manufactured based on short-term forecasts. Planning of the two subsystems using a two-phase sequential approach offers both operational and modeling incentives. In this paper, we formulate a two-phase mixed integer linear programming (MILP) model for material procurement, components fabrication, product assembly and distribution scheduling of a BTO supply chain system. In the proposed approach, the entire problem is first decomposed into two subsystems and evaluated sequentially. The first phase deals with assembling and distribution scheduling of customizable products, while the second phase addresses fabrication and procurement planning of components and raw-materials. The objective of both models is to minimize the aggregate costs associated with each subsystem, while meeting customer service requirements. The search space for the first phase problem involves a complex landscape with too many candidate solutions. A genetic algorithm based solution procedure is proposed to solve the sub-problem efficiently.

Autonomous parallel parking of a car-like mobile robot by a neuro-fuzzy sensor-based controller

In this paper, the concept of sensor-based behavior is used to design a neuro-fuzzy control system for a car-like-mobile-robot. The problem addressed is the parallel parking in a rectangular constrained space with just one backward maneuver. To accomplish the autonomous fuzzy behavior control, the car-like-mobile-robot has trained to park in just 2 parking dimensions based on the training data obtained from sensor information generated offline by adopting a fifth-order polynomial as the reference trajectory. The proposed controller is an ANFIS architecture that generates turning angle as output. As long as the states (positions and orientations) of the robot are measurable at each discrete-time step during the control process, this controller can make the robot follow feasible trajectories by just knowing the initial configuration of the robot and park successfully at the prescribed goal position. The simulation results which are based on real dimensions of a typical car demonstrate the feasibility and effectiveness of the proposed controller in practical car maneuvers.

Fuzzy scheduling of a build-to-order supply chain

The overwhelming majority of the literature in the area of supply chain planning and scheduling considers the traditional make-to-stock (MTS) environment. However, manufacturers of assembled products such as cars, computers, furniture, etc. adopt the build-to-order supply chain (BOSC) to become agile in a mass customization process in order to meet diversified customer requirements. In this paper we propose an integrated production–distribution planning model for a multi-echelon, multi-plant and multi-product supply chain operating in a build-to-order (BTO) environment. The uncertainties associated with estimation of the various operational cost parameters are represented by fuzzy numbers. The BOSC scheduling model is thus constructed as a mixed-integer fuzzy programming (MIFP) problem with the goal of reducing the overall operating costs related to component fabrication, procurement, assembling, inspection, logistics and inventory, while improving customer satisfaction by allowing product customization and meeting delivery promise dates at each market outlet. An efficient compromise solution approach by transforming the problem into an auxiliary multi-objective linear programming model is also suggested.

Sonar based mobile robot localization by using fuzzy triangulation

Abstract The objective of this paper is to identify the robot’s location in a global map from solely sonar based information. This is achieved by using fuzzy sets to model sonar data and by using fuzzy triangulation to identify robot’s position and orientation. As a result we obtain a fuzzy position region where each point in the region has a degree of certainty of being the actual position of the robot.

A hierarchical neuro-fuzzy system to near optimal-time trajectory planning of redundant manipulators

In this paper, the problem of minimum-time trajectory planning is studied for a three degrees-of-freedom planar manipulator using a hierarchical hybrid neuro-fuzzy system. A first neuro-fuzzy network named NeFIK is considered to solve the inverse kinematics problem. After a few pre-processing steps characterizing the minimum-time trajectory and the corresponding torques, a second neuro-fuzzy controller is built. Its purpose is to fit the robot dynamic behavior corresponding to the determined minimum-time trajectory with respect to actuators models, torque nominal values, as well as position, velocity, acceleration and jerk boundary conditions. A Tsukamoto Neuro-Fuzzy Inference network is designed to achieve the online control of the robot. The premise parameters (antecedent membership functions parameters) as well as rule-consequence parameters are learned and optimized, generating the optimal-time trajectory torques, representing the robot dynamic behavior. Simulation results are presented and discussed.

Autonomous parallel parking of a car-like mobile robot by a neuro-fuzzy behavior-based controller

In this paper, the concept of sensor-based behavior is used to design a neuro-fuzzy control system for a car-like-mobile-robot. The problem addressed is the parallel parking in a rectangular constrained space with just one backward maneuver. To accomplish the autonomous fuzzy behavior control, the car-like-mobile-robot has trained to park in just 2 parking dimensions based on the training data obtained from sensor information generated offline by adopting a fifth-order polynomial as the reference trajectory. The proposed controller is an ANFIS architecture that generates turning angle as output. As long as the states (positions and orientations) of the robot are measurable at each discrete-time step during the control process, this controller can make the robot follow feasible trajectories by just knowing the initial configuration of the robot and park successfully at the prescribed goal position. The simulation results which are based on real dimensions of a typical car demonstrate the feasibility and effectiveness of the proposed controller in practical car maneuvers.

Fuzzy dynamic localization for mobile robots

In this paper, we introduce a new fuzzy logic-based approach for dynamic localization of mobile robots equipped with a ring of sonar sensors. In this approach, the angular uncertainty and radial imprecision of sonar data are modeled by possibility distributions. From sonar data, a local fuzzy composite map is constructed and fitted to the given global map of the environment to identify robots location. As a result of this fit, either a unique fuzzy location or multiple candidate fuzzy locations are obtained. To reduce the multiple candidate locations, the robot is moved to a new location and a new local fuzzy composite map is constructed. Then, a new set of candidate fuzzy locations is obtained. By considering the robots movement, a set of hypothesized locations is identified from the old set of candidate locations. The hypothesized locations are matched with the new candidate locations and the candidates with low degree of match are eliminated. This process is continued until a unique location is obtained. The matching process is performed by using the fuzzy pattern matching technique. The proposed method is implemented on a Nomad 200 robot and the results are reported.

Adaptive control of a class of nonlinear systems with a first-order parameterized Sugeno fuzzy approximator

In this paper, an adaptive fuzzy control scheme for tracking a class of continuous-time plants is presented. A parameterized Sugeno fuzzy approximator is used to adaptively compensate for the plant nonlinearities. All parameters in the fuzzy approximator are tuned using a Lyapunov-based design. In the fuzzy approximator, first-order Sugeno consequents are used in the IF-THEN rules of the fuzzy system, which has a better approximation capability than those using constant consequents. Global boundedness of the adaptive system is established. Finally, a simulation is used to demonstrate the effectiveness of the proposed controller.

Fuzzy scheduling of job orders in a two-stage flowshop with batch-processing machines

In this paper, we present a mixed-integer fuzzy programming model and a genetic algorithm (GA) based solution approach to a scheduling problem of customer orders in a mass customizing furniture industry. Independent job orders are grouped into multiple classes based on similarity in style so that the required number of setups is minimized. The family of jobs can be partitioned into batches, where each batch consists of a set of consecutively processed jobs from the same class. If a batch is assigned to one of available parallel machines, a setup is required at the beginning of the first job in that batch. A schedule defines the way how the batches are created from the independent jobs and specifies the processing order of the batches and that of the jobs within the batches. A machine can only process one job at a time, and cannot perform any processing while undergoing a setup. The proposed formulation minimizes the total weighted flowtime while fulfilling due date requirements. The imprecision associated with estimation of setup and processing times are represented by fuzzy sets.