Christoph Greulich

Solving Physical Traveling Salesman Problems with policy adaptation

The Physical Traveling Salesman Problem (PTSP) is a current research problem which adds a model of velocity to the classic TSP. In this paper we propose algorithms for solving the PTSP which avoid the fragmented allocation of memory and precompute cell-precise single-source shortest paths for each waypoint by using an engineered implementation of Dijkstras algorithm. To determine an initial tour, we solve ordinary and general TSPs. For moderately sized problems, we apply an optimal depth-first branch-and-bound TSP solver which warrants constant-time per search tree node. For larger problems, we apply randomized search with policy adaptation to learn from good tours. We evaluate our solution with a series of benchmark experiments and compare the results to the winner of the PTSP competition at CIG 2013. In comparison, our approach shows similar results but also provides a graph search with optimal time performance.

Monte-Carlo Tree Search for Logistics

In this paper we review recent advances of randomized AI search in solving industrially relevant optimization problems. The method we focus on is a sampling-based solution mechanism called Monte-Carlo Tree Search (MCTS), which is extended by the concepts of nestedness and policy adaptation to establish a better trade-off between exploitation and exploration. This method, originating in game playing research, is a general heuristic search technique, for which often less problem-specific knowledge has to be added than in comparable approaches.

Agent-Based Multimodal Transport Planning in Dynamic Environments

The development and maintenance of traffic concepts in urban districts is expensive and leads to high investments for altering transport infrastructures or for the acquisition of new resources. We present an agent-based approach for modeling, simulation, evaluation, and optimization of public transport systems by introducing a dynamic microscopic model. Actors of varying stakeholders are represented by intelligent agents. While describing the inter-agent communication and their individual behaviors, the focus is on the implementation of information systems for traveler agents as well as on the matching between open source geographic information systems, and standardized transport schedules provided by the Association of German Transport Companies. The performance, efficiency, and limitations of the system are evaluated within the public transport infrastructure of Bremen. We discuss the effects of passengers’ behaviors to the entire transport network and investigate the system’s flexibility as well as consequences of incidents in travel plans.

Cyber-Physical Multiagent-Simulation in Production Logistics

A growing network of technical systems, embedded and autonomous, influence our daily work. Among them, cyber-physical systems establish a close connection between the virtual and the real world. In this paper we show how an existing multiagent system that controls the physical production of goods on a monorail is virtualized by extracting the agents as black boxes and by integrating them into a multiagent simulation system. As a result, the exact same agents run in physical and cyber world. Towards this end, the physical environment has been mapped and visualized. Experiments show that the modeling and simulation error is small, such that scenarios can be varied, tested, debugged, and scaled, saving huge amounts of labor.

Optimizing Last Mile Delivery Using Public Transport with Multi-Agent Based Control

The majority of research work carried out in the field of Operations Research have relied on optimization algorithms to improve the pick-up and delivery problem. Most studies aim to solve the vehicle routing problem, to accommodate optimum delivery orders, vehicles etc. This paper focuses on developing a system model, which uses existing Public Transport facility of a city for the transportation of small and medium sized packaged goods, to avoid further aggravating the situation of urban congestion and also help reduce green house gas emissions. The research carried out investigates the feasibility of the proposed multi-agent based simulation model, for efficiency of cost, time and energy consumption. The Dijkstra Shortest Path algorithm and Nested Monte Carlo Search have been implemented to build a time based cost matrix which is used to generate a tour plan for intermodal delivery of goods. The quality of the tour is dependent on the efficiency of the search algorithm implemented for plan generation and route planning. The results reveal a definite advantage of using Public Transportation over existing delivery approaches in terms of energy efficiency.

Using SPIN for the Optimized Scheduling of Discrete Event Systems in Manufacturing

A discrete event system (DES) is a dynamic system with discrete states the transitions of which are triggered by events. In this paper we propose the application of the Spin software model checker to a discrete event system that controls the industrial production of autonomous products. The flow of material is asynchronous and buffered. The aim of this work is to find concurrent plans that optimize the throughput of the system. In the mapping the discrete event system directly to the model checker, we model the production line as a set of communicating processes, with the movement of items modeled as channels. Experiments shows that the model checker is able to analyze the DES, subject to the partial ordering of the product parts. It derives valid and optimized plans with several thousands of steps using constraint branch-and-bound.

A case study of planning for smart factories: Model checking and Monte Carlo search for the rescue

In this work, we propose the application of the SPIN software model checker to a multiagent system that controls the industrial production of goods. The flow of material is buffered on a production line with assembling stations. As the flow of material is asynchronous at each station, queuing is required as long as buffers provide waiting room. Besides validating the design of the system, the core objective of this work is to find concurrent plans that optimize the throughput of the system. In the mapping of the production system to the model checker, we model the production line as a set of communicating processes, with the movement of items modeled as channels. Experiments show that the model checker is able to analyze the system, subject to the partial ordering of the product parts. It derives valid and optimized plans with several thousands of steps using constraint branching in branch-and-bound search. We compare the results with a randomized exploration based on recent advances in Monte Carlo search.

Branch-and-Bound Optimization of a Multiagent System for Flow Production using Model Checking

In this paper we propose the application of a model checker to evaluate a multiagent system that controls the industrial production of autonomous products. As the flow of material is asynchronous at each station, queuing effects arise as long as buffers provide waiting room. Besides validating the design of the system, the core objective of this work is to find plans that optimize the throughput of the system. Instead of mapping the multiagent system directly to the model checker, we model the production line as a set of communicating processes, with the movement of items modeled as communication channels. Experiments shows that the model checker is able to analyze the movements of autonomous products for the model, subject to the partial ordering of the product parts. It derives valid and optimized plans with several thousands of steps using constraint branch-and-bound.

Nested Rollout Policy Adaptation for Multiagent System Optimization in Manufacturing.

In manufacturing there are not only flow lines with stations arranged one behind the other, but also more complex networks of stations where assembly operations are performed. The considerable difference from sequential flow lines is that a partially ordered set of required components are brought together in order to form a single unit at the assembly stations in a competitive multiagent system scenario. In this paper we optimize multiagent control for such flow production units with recent advances of Nested Monte-Carlo Search. The optimization problem is implemented as a single-agent game in a generic search framework. In particular, we employ Nested Monte-Carlo Search with Rollout Policy Adaptation and apply it to a modern flow production unit, comparing it to solutions obtained with a simulator and with a model checker.

Two Model Checking Approaches to Branch-and-Bound Optimization of a Flow Production System

In this paper we introduce a novel application of model checking to find optimal planning solutions for a flow production system. Originally controlled by a multiagent system, the production system consists of autonomous products and asynchronous production stations with limited space for waiting products. In this work, we present two different approaches of application of the Spin model checker to optimize throughput in the given production system. Instead of mapping the multiagent system directly, we model the production line itself as a set of communicating processes. Each communication channel between two processes represents a one-way monorail connection from one station to another. Experiments show that both approaches derive valid and optimized plans with several thousands of steps using constrained branch-and-bound. However, experiments also indicate individual advantages of both approaches.