Stefan Vonolfen

Architecture and Design of the HeuristicLab Optimization Environment

Many optimization problems cannot be solved by classical mathematical optimization techniques due to their complexity and the size of the solution space. In order to achieve solutions of high quality though, heuristic optimization algorithms are frequently used. These algorithms do not claim to find global optimal solutions, but offer a reasonable tradeoff between runtime and solution quality and are therefore especially suitable for practical applications. In the last decades the success of heuristic optimization techniques in many different problem domains encouraged the development of a broad variety of optimization paradigms which often use natural processes as a source of inspiration (as for example evolutionary algorithms, simulated annealing, or ant colony optimization). For the development and application of heuristic optimization algorithms in science and industry, mature, flexible and usable software systems are required. These systems have to support scientists in the development of new algorithms and should also enable users to apply different optimization methods on specific problems easily. The architecture and design of such heuristic optimization software systems impose many challenges on developers due to the diversity of algorithms and problems as well as the heterogeneous requirements of the different user groups. In this chapter the authors describe the architecture and design of their optimization environment HeuristicLab which aims to provide a comprehensive system for algorithm development, testing, analysis and generally the application of heuristic optimization methods on complex problems.

Simulation-based evolution of resupply and routing policies in rich vendor-managed inventory scenarios

Vendor managed inventory combines inventory management and transportation. Compared to classical inventory management approaches, this strategy offers various degrees of freedom for the vendor while providing a certain service quality level for the customers. To capture the characteristics of rich real-world scenarios, our problem formulation consists of multiple customers, many products and stochastic product usages. Additionally, we also consider mixed formulations, where only a certain part of the customers is switched to a vendor managed inventory to allow a stepwise transition. We show that resupply and routing policies can be evolved autonomously for those scenarios using a simulation-based optimization approach. By combining inventory management and routing, the resulting policies aim to minimize costs and to maximize resource usage while maintaining a given service level. In order to validate our approach, we perform case studies and apply the evolved rules on a large-scale vendor managed inventory scenario for supermarkets. Furthermore, we show that our methodology can be used to perform a sensitivity analysis by considering the influence of exogenous and endogenous factors on the decision process, if a customer base should be transitioned to a vendor managed inventory.

Distribution of waiting time for dynamic pickup and delivery problems

Pickup and delivery problems have numerous applications in practice such as parcel delivery and passenger transportation. In the dynamic variant of the problem, not all information is available in advance but is revealed during the planning process. Thus, it is crucial to anticipate future events in order to generate high-quality solutions. Previous work has shown that the use of waiting strategies has the potential to save costs and maximize service quality. We adapt various waiting heuristics to the pickup and delivery problem with time windows. Previous research has shown, that specialized waiting heuristics utilizing anticipatory knowledge potentially outperform general heuristics. Direct policy search based on evolutionary computation and a simulation model is proposed as a methodology to automatically specialize waiting strategies to different problem characteristics. Based on the strengths of the previously introduced waiting strategies, we propose a novel waiting heuristic that can utilize historical request information based on an intensity measure which does not require an additional data preprocessing step. The performance of the waiting heuristics is evaluated on a single set of benchmark instances containing various instance classes that differ in terms of spatial and temporal properties. The diverse set of benchmark instances is used to analyze the influence of spatial and temporal instance properties as well as the degree of dynamism to the potential savings that can be achieved by anticipatory waiting and the incorporation of knowledge about future requests.

Simulation-Based Optimization with HeuristicLab: Practical Guidelines and Real-World Applications

Dynamic and stochastic problem environments are often difficult to model using standard problem formulations and algorithms. One way to model and then solve them is simulation-based optimization: Simulations are integrated into the optimization process in order to evaluate the quality of solution candidates and to identify optimized system configurations. Potential solutions are evaluated with a simulation model, which leads to new challenges regarding runtime performance, robustness, and distributed evaluation. In order to design, compare, and parameterize algorithmic approaches it is beneficial to use an optimization framework for algorithm design and evaluation. On the one hand, this chapter shows how arbitrary simulators can be coupled with the open-source HeuristicLab optimization framework. This coupling is implemented in a generic way so that the simulators act as external evaluators. On the other hand, we demonstrate how arbitrary optimizers available within HeuristicLab can be called from a simulator in order to perform complex optimization tasks within the simulation model. In order to illustrate the applicability of these approaches, real-world examples investigated by the authors are discussed. We show here application examples from different fields, namely logistics network design, vendor managed inventory routing, steel slab logistics, production optimization with dispatching rule scheduling, material flow simulation, and layout optimization.

Structural Synthesis of Dispatching Rules for Dynamic Dial-a-Ride Problems

The dial-a-ride problem consists of designing vehicle routes in the area of passenger transportation. Assuming that each vehicle can act autonomously, the problem can be modeled as a multi-agent system. In that context, it is a complex decision process for each agent to determine what action to perform next. In this work, the agent function is evolved using genetic programming by synthesizing basic bits of information. Specialized dispatching rules are synthesized automatically that are adapted to the problem environment. We compare the evolved rules with other dispatching strategies for dynamic dial-a-ride problems on a set of generated benchmark instances. Additionally, since genetic programming is a whitebox-based approach, insights can be gained about important system parameters. For that purpose, we perform a variable frequency analysis during the evolutionary process.

Optimizing assembly line supply by integrating warehouse picking and forklift routing using simulation

The significance of system orientation in production and logistics optimization has often been neglected in the past. An isolated view on single activities may result in globally suboptimal performance. We consider a manufacturing process where assembly lines are supplied from a central logistics center. The different steps, such as storage, picking and transport of work-in-process materials to and from the assembly lines, strongly influence each other. For instance, if the picking process batches orders that need to be transported to the same target, a reduction of travel distances can be achieved. The individual problems are coupled and validated via simulation, which leads to more robust and applicable results in practice. We test our approach on a scenario based on real-world data from Rosenbauer, one of the worlds largest suppliers of firefighting vehicles. Our results indicate that warehouse optimization can lead to a more efficient transport in an integrated problem formulation.

Simulation-based evolution of municipal glass-waste collection strategies utilizing electric trucks

In urban areas freight transport imposes many social and environmental issues. Solid waste collection accounts for a considerable amount of freight transportation in municipal areas. We propose a simulation-based approach to evaluate different scenarios where electric trucks replace conventional trucks for the collection of municipal glass-waste. Under special consideration of the characteristics of electric trucks, the simulation of real-time fill level information is used and coupled with an optimization component to evolve adapted waste collection strategies. We illustrate our approach on two test-scenarios based on real-world data. We use the simulation model to evaluate several scenarios in terms of costs and environmental impact and the optimization environment to generate collection strategies that minimize those performance figures while maintaining a given service quality.

Solving large-scale vehicle routing problem instances using an island-model offspring selection genetic algorithm

The vehicle routing problem is a class of problems that frequently occurs in the field of transportation logistics. In this work, we tackle very-large scale problem instances with time windows. Among other techniques, metaheuristics are frequently used to solve large-scale instances close to optimality. We present an island-model genetic algorithm variant and apply several techniques such as offspring selection and adaptive constraint relaxation. To validate our approach, we perform test runs on benchmark instances with 1000 customers and compare the results to the currently best-known solutions.

Genetic programming enabled evolution of control policies for dynamic stochastic optimal power flow

The optimal power flow (OPF) is one of the central optimization problems in power grid engineering, building an essential tool for numerous control as well as planning issues. Methods for solving the OPF that mainly treat steady-state situations have been studied extensively, ignoring uncertainties of system variables as well as their volatile behavior. While both the economical as well as well as technical importance of accurate control is high, especially for power flow control in dynamic and uncertain power systems, methods are needed that provide (near-) optimal actions quickly, eliminating issues on convergence speed or robustness of the optimization. This paper shows an approximate policy-based control approach where optimal actions are derived from policies that are learned offline, but that later provide quick and accurate control actions in volatile situations. These policies are evolved using genetic programming, where multiple and interdependent policies are learned synchronously with simulation-based optimization. Finally, an approach is available for learning fast and robust power flow control policies suitable to highly dynamic power systems such as smart electric grids.

Modelling and optimizing storage assignment in a steel slab yard

Steel slabs are intermediates in the production of sheets, plates or coils in the steel industry. In this paper we consider cold charge slabs which are stored in stacks on a slab yard for a couple of hours, days or weeks until they are assigned to a rolling schedule and retrieved. When a slab is not positioned on top of a stack, retrieval requires the movement, also called shuffling, of all slabs above. We introduce a new model formulation, suitable neighbourhood and perturbation operators and discuss strategies for optimizing slab yard assignments with respect to shuffles and travel distance in picking.