Giacomo Da Col | Alpen-Adria-Universität Klagenfurt (original) (raw)
Papers by Giacomo Da Col
Intelligent Computing. CompCom 2019. Advances in Intelligent Systems and Computing, 2019
Games represent important benchmark problems for AI. One-player games, also called puzzles, often... more Games represent important benchmark problems for AI. One-player games, also called puzzles, often resemble real world optimization problems and, thus, lessons learned on such games are also important for such problems. In this paper we focus on the game of Tetris, which can also be seen as a packing problem variant. We provide an empirical evaluation of a heuristic search approach for Tetris with the following goal: Having an effective heuristic function at hand, we want to answer the question how much the additional tree search pays off. We are especially interested if there is a so called sweet spot that represents the best ratio between score achieved and time invested in the search. This knowledge is crucial in order to be able to implement deep-learning approaches in the light of limited computing resources, i.e. to produce many good games to be learned from in rather short time. Our experiments reveal that such a sweet spot exists and hence, using this knowledge, only a fraction of time is needed for producing the same amount of learning data with similar quality.
Proceedings of the 8th International Workshop on Combinations of Intelligent Methods and Applications co-located with 30th International Conference on Artificial Intelligence Tools (ICTAI 2018), 2018
Generating optimized large-scale production plans is an important open problem where even small i... more Generating optimized large-scale production plans is an important open problem where even small improvements result in significant savings. Application scenarios in the semiconductor industry comprise thousands of machines and hundred thousands of job operations and are therefore among the most challenging scheduling problems regarding their size. In this paper we present a novel approach for automatically creating composite dispatching rules, i.e. heuristics for job sequencing, for makespan optimization in such large-scale job shops. The approach builds on the combination of event-based simulation and genetic algorithms. We introduce a new set of benchmark instances with proven optima that comprise up to 100000 operations to be scheduled on up to 1000 machines. With respect to this large-scale benchmark, we present the results of an experiment comparing well-known dispatching rules with automatically created composite dispatching rules produced by our system. It is shown that the proposed system is able to come up with highly effective dispatching rules such that makespan reductions of up to 38% can be achieved, and in fact, often near optimal or even optimal schedules can be produced.
Proceedings of the 19th International Configuration Workshop, 2017
In this paper, we propose an approach for learning heuristics for constraint satisfaction problem... more In this paper, we propose an approach for learning heuristics for constraint satisfaction problems in general and for configuration problems in particular. The genetic algorithm based learning approach automatically derives variable ordering, value ordering and pruning strategies for the exploitation by constraint solvers. We evaluate our approach with respect to the combined configuration problem, which is a generic configuration problem including sub problems such as graph coloring or bin packing. The results show that one of the best performing heuristics identified by our approach performs equally well compared to the expert heuristic defined in cooperation with our project partners from Siemens.
KI 2016: Advances in Artificial Intelligence , 2016
Job-shop scheduling problems constitute a big challenge in nowadays industrial manufacturing envi... more Job-shop scheduling problems constitute a big challenge in nowadays industrial manufacturing environments. Because of the size of realistic problem instances, applied methods can only afford low computational costs. Furthermore, because of highly dynamic production regimes, adaptability is an absolute must. In state-of-the-art production factories the large-scale problem instances are split into subinstances, and greedy dispatching rules are applied to decide which job operation is to be loaded next on a machine. In this paper we propose a novel scheduling approach inspired by those hand-crafted scheduling routines. Our approach builds on problem decomposition for keeping computational costs low, dispatching rules for effectiveness and declarative programming for high adaptability and maintainability. We present first results proving the concept of our novel scheduling approach based on a new large-scale job-shop benchmark with proven optimal solutions.
Proceedings of Austrian Robotics Workshop, 2015
In explorations it is often required for mobile robotic explorers to update a base station about ... more In explorations it is often required for mobile
robotic explorers to update a base station about the progress of
the mission. Robots may form a mobile ad hoc network to establish
connectivity. To connect mobile robots and base stations
two strategies have been suggested: multi-hop and rendezvous.
For multi-hop connections mobile robots form a daisy chain
propagating data between both ends. In rendezvous, robots
carry data by driving between communicating entities. While
both strategies have been implemented and tested, no formal
comparison of both strategies has been done. We determine
various parameters having an impact on the exploration time of
both strategies. Using a mathematical model we quantitatively
evaluate their impact. Better understanding of the parameters
allows to optimize both strategies for a given application.
A general assertion whether rendezvous or multi-hop yields
shorter exploration times cannot be made.
Intelligent Computing. CompCom 2019. Advances in Intelligent Systems and Computing, 2019
Games represent important benchmark problems for AI. One-player games, also called puzzles, often... more Games represent important benchmark problems for AI. One-player games, also called puzzles, often resemble real world optimization problems and, thus, lessons learned on such games are also important for such problems. In this paper we focus on the game of Tetris, which can also be seen as a packing problem variant. We provide an empirical evaluation of a heuristic search approach for Tetris with the following goal: Having an effective heuristic function at hand, we want to answer the question how much the additional tree search pays off. We are especially interested if there is a so called sweet spot that represents the best ratio between score achieved and time invested in the search. This knowledge is crucial in order to be able to implement deep-learning approaches in the light of limited computing resources, i.e. to produce many good games to be learned from in rather short time. Our experiments reveal that such a sweet spot exists and hence, using this knowledge, only a fraction of time is needed for producing the same amount of learning data with similar quality.
Proceedings of the 8th International Workshop on Combinations of Intelligent Methods and Applications co-located with 30th International Conference on Artificial Intelligence Tools (ICTAI 2018), 2018
Generating optimized large-scale production plans is an important open problem where even small i... more Generating optimized large-scale production plans is an important open problem where even small improvements result in significant savings. Application scenarios in the semiconductor industry comprise thousands of machines and hundred thousands of job operations and are therefore among the most challenging scheduling problems regarding their size. In this paper we present a novel approach for automatically creating composite dispatching rules, i.e. heuristics for job sequencing, for makespan optimization in such large-scale job shops. The approach builds on the combination of event-based simulation and genetic algorithms. We introduce a new set of benchmark instances with proven optima that comprise up to 100000 operations to be scheduled on up to 1000 machines. With respect to this large-scale benchmark, we present the results of an experiment comparing well-known dispatching rules with automatically created composite dispatching rules produced by our system. It is shown that the proposed system is able to come up with highly effective dispatching rules such that makespan reductions of up to 38% can be achieved, and in fact, often near optimal or even optimal schedules can be produced.
Proceedings of the 19th International Configuration Workshop, 2017
In this paper, we propose an approach for learning heuristics for constraint satisfaction problem... more In this paper, we propose an approach for learning heuristics for constraint satisfaction problems in general and for configuration problems in particular. The genetic algorithm based learning approach automatically derives variable ordering, value ordering and pruning strategies for the exploitation by constraint solvers. We evaluate our approach with respect to the combined configuration problem, which is a generic configuration problem including sub problems such as graph coloring or bin packing. The results show that one of the best performing heuristics identified by our approach performs equally well compared to the expert heuristic defined in cooperation with our project partners from Siemens.
KI 2016: Advances in Artificial Intelligence , 2016
Job-shop scheduling problems constitute a big challenge in nowadays industrial manufacturing envi... more Job-shop scheduling problems constitute a big challenge in nowadays industrial manufacturing environments. Because of the size of realistic problem instances, applied methods can only afford low computational costs. Furthermore, because of highly dynamic production regimes, adaptability is an absolute must. In state-of-the-art production factories the large-scale problem instances are split into subinstances, and greedy dispatching rules are applied to decide which job operation is to be loaded next on a machine. In this paper we propose a novel scheduling approach inspired by those hand-crafted scheduling routines. Our approach builds on problem decomposition for keeping computational costs low, dispatching rules for effectiveness and declarative programming for high adaptability and maintainability. We present first results proving the concept of our novel scheduling approach based on a new large-scale job-shop benchmark with proven optimal solutions.
Proceedings of Austrian Robotics Workshop, 2015
In explorations it is often required for mobile robotic explorers to update a base station about ... more In explorations it is often required for mobile
robotic explorers to update a base station about the progress of
the mission. Robots may form a mobile ad hoc network to establish
connectivity. To connect mobile robots and base stations
two strategies have been suggested: multi-hop and rendezvous.
For multi-hop connections mobile robots form a daisy chain
propagating data between both ends. In rendezvous, robots
carry data by driving between communicating entities. While
both strategies have been implemented and tested, no formal
comparison of both strategies has been done. We determine
various parameters having an impact on the exploration time of
both strategies. Using a mathematical model we quantitatively
evaluate their impact. Better understanding of the parameters
allows to optimize both strategies for a given application.
A general assertion whether rendezvous or multi-hop yields
shorter exploration times cannot be made.