Juho Andelmin - Academia.edu (original) (raw)
Papers by Juho Andelmin
With this release, we're past initial development, and the package should be relatively stabl... more With this release, we're past initial development, and the package should be relatively stable. Over the past few months, we have worked on improving both usability and documentation and making the interface more user-friendly. Combined with earlier improvements to the model (v.0.2.0), the package is ready for use.
Influence diagrams are widely employed to represent the structure of discrete multi-stage decisio... more Influence diagrams are widely employed to represent the structure of discrete multi-stage decision problems under uncertainty. In this paper, we develop the Decision Programming framework which extends the capabilities of influence diagrams through formulations that permit the modeling of many kinds of constraints so that optimal solutions can be established with mixed-integer linear programming. In particular, Decision Programming makes it possible to (i) omit the usual `no forgetting' assumption; (ii) accommodate both deterministic and chance constraints, including those based on risk measures such as Conditional Value-at-Risk; and (iii) determine all non-dominated strategies in the case of multiple value objectives. In the context of project portfolio selection, Decision Programming can be viewed as an extension of Contingent Portfolio Programming to problems in which scenario probabilities depend endogenously on project decisions. We provide illustrative examples and evidenc...
European Journal of Operational Research, 2020
Transportation Science, 2017
We propose an exact algorithm for solving the green vehicle routing problem (G-VRP). The G-VRP mo... more We propose an exact algorithm for solving the green vehicle routing problem (G-VRP). The G-VRP models the optimal routing of an alternative fuel vehicle fleet to serve a set of geographically scattered customers. Vehicles’ fuel autonomy and possible refueling stops en route are explicitly modeled and maximum duration constraints are imposed on each vehicle route. We model the G-VRP as a set partitioning problem in which columns represent feasible routes corresponding to simple circuits in a multigraph. Each node in the multigraph represents one customer and each arc between two customers represents a nondominated path through a set of refueling stations visited by a vehicle when traveling directly between the two customers. We strengthen the set partitioning formulation by adding valid inequalities including k-path cuts and describe a method for separating them. We provide computational results on benchmark instances showing that the algorithm can optimally solve instances with up to ∼110 customers. The o...
arXiv: Optimization and Control, 2019
Influence diagrams are widely employed to represent multi-stage decision problems in which each d... more Influence diagrams are widely employed to represent multi-stage decision problems in which each decision is a choice from a discrete set of alternatives, uncertain chance events have discrete outcomes, and prior decisions may influence the probability distributions of uncertain chance events endogenously. In this paper, we develop the Decision Programming framework which extends the applicability of influence diagrams by developing mixed-integer linear programming formulations for solving such problems in the presence of many kinds of constraints. In particular, Decision Programming makes it possible to (i) solve problems in which earlier decisions cannot necessarily be recalled later, for instance, when decisions are taken by agents who cannot communicate with each other; (ii) accommodate a broad range of deterministic and chance constraints, including those based on resource consumption, logical dependencies or risk measures such as Conditional Value-at-Risk; and (iii) determine a...
This document can be stored and made available to the public on the open internet pages of Aalto ... more This document can be stored and made available to the public on the open internet pages of Aalto University. All other rights are reserved.
vehicles (BEVs) is studied. The studied problem involves a fleet of identical BEVs located at a c... more vehicles (BEVs) is studied. The studied problem involves a fleet of identical BEVs located at a central depot, a set of customers that must be serviced within given time windows, and a set of charging stations where the vehicles can recharge their batteries. The objective is to design a set of vehicle routes, each starting and ending at the depot, so that each customer is serviced exactly once and the total energy cost of the vehicle routes is minimized. Since BEVs have limited battery capacity and low recharging rate, charging station visits and recharging times must be explicitly considered in the route planning, wherefore most VRP variants are not sufficient in modeling the studied problem. Optimal routing of electric vehicles is not much studied in the optimization literature. The two most relevant models are the Green Vehicle Routing Problem (G-VRP) by Erdoğan & Miller-Hooks (2012) and the Electric VRP with Time Windows (E-VRPTW) by Schneider et al. (2013). The model considered in this thesis generalizes the E-VRPTW by also allowing the possibility of recharging a variable amount of energy at charging stations (variable recharging scheme) rather than performing a full recharge at every visit (fixed recharging scheme). This thesis introduces energy paths (e-paths) and proposes a new formulation of the studied problem based on non-dominated e-paths between every customer pair. The new formulation reduces the number of decision variables in the model and eliminates the need of imposing an artificial upper bound on the number of stops to a charging station, as is commonly done in previous models to keep their size acceptable. Some new preprocessing steps and valid inequalities are also presented to strengthen the LP-relaxation of the proposed formulation. Computational tests indicate that the new formulation provides considerable improvements over the standard formulation. Moreover, it is shown that significant reductions in the routing cost can be obtained in real-world routing problems by adopting the variable recharging scheme over the fixed one.
Influence diagrams are widely employed to represent multi-stage decision problems in which each d... more Influence diagrams are widely employed to represent multi-stage decision problems in which each decision is a choice from a discrete set of alternatives, uncertain chance events have discrete outcomes, and prior decisions may influence the probability distributions of uncertain chance events endogenously. In this paper, we develop the Decision Programming framework which extends the applicability of influence diagrams by developing mixedinteger linear programming formulations for solving such problems in the presence of many kinds of constraints. In particular, Decision Programming makes it possible to (i) solve problems in which earlier decisions cannot necessarily be recalled later, for instance, when decisions are taken by agents who cannot communicate with each other; (ii) accommodate a broad range of deterministic and chance constraints, including those based on resource consumption, logical dependencies or risk measures such as Conditional Value-at-Risk; and (iii) determine al...
Computers & Operations Research
Abstract We consider the Green Vehicle Routing Problem (G-VRP) which is an extension of the class... more Abstract We consider the Green Vehicle Routing Problem (G-VRP) which is an extension of the classical vehicle routing problem for alternative fuel vehicles. In the G-VRP, vehicles’ driving autonomy and possible refueling stops en-route are explicitly modeled. We propose a multi-start local search algorithm that consists of three phases. The first two phases iteratively construct new solutions, improve them by local search, and store all vehicle routes forming these solutions in a route pool. Phase three optimally combines vehicle routes in the route pool by solving a set partitioning problem and improves the final solution by local search. The algorithm is based on a multigraph reformulation of the G-VRP in which nodes correspond to customers and a depot, and arcs correspond to possible sequences of refueling stops for vehicles traveling between two nodes. All local search operators used by our algorithm are tailored to exploit this reformulation and do not explicitly deal with refueling stations. We report computational results on benchmark instances with up to ∼ 470 customers, showing that the algorithm is competitive with state-of-the-art heuristics.
With this release, we're past initial development, and the package should be relatively stabl... more With this release, we're past initial development, and the package should be relatively stable. Over the past few months, we have worked on improving both usability and documentation and making the interface more user-friendly. Combined with earlier improvements to the model (v.0.2.0), the package is ready for use.
Influence diagrams are widely employed to represent the structure of discrete multi-stage decisio... more Influence diagrams are widely employed to represent the structure of discrete multi-stage decision problems under uncertainty. In this paper, we develop the Decision Programming framework which extends the capabilities of influence diagrams through formulations that permit the modeling of many kinds of constraints so that optimal solutions can be established with mixed-integer linear programming. In particular, Decision Programming makes it possible to (i) omit the usual `no forgetting' assumption; (ii) accommodate both deterministic and chance constraints, including those based on risk measures such as Conditional Value-at-Risk; and (iii) determine all non-dominated strategies in the case of multiple value objectives. In the context of project portfolio selection, Decision Programming can be viewed as an extension of Contingent Portfolio Programming to problems in which scenario probabilities depend endogenously on project decisions. We provide illustrative examples and evidenc...
European Journal of Operational Research, 2020
Transportation Science, 2017
We propose an exact algorithm for solving the green vehicle routing problem (G-VRP). The G-VRP mo... more We propose an exact algorithm for solving the green vehicle routing problem (G-VRP). The G-VRP models the optimal routing of an alternative fuel vehicle fleet to serve a set of geographically scattered customers. Vehicles’ fuel autonomy and possible refueling stops en route are explicitly modeled and maximum duration constraints are imposed on each vehicle route. We model the G-VRP as a set partitioning problem in which columns represent feasible routes corresponding to simple circuits in a multigraph. Each node in the multigraph represents one customer and each arc between two customers represents a nondominated path through a set of refueling stations visited by a vehicle when traveling directly between the two customers. We strengthen the set partitioning formulation by adding valid inequalities including k-path cuts and describe a method for separating them. We provide computational results on benchmark instances showing that the algorithm can optimally solve instances with up to ∼110 customers. The o...
arXiv: Optimization and Control, 2019
Influence diagrams are widely employed to represent multi-stage decision problems in which each d... more Influence diagrams are widely employed to represent multi-stage decision problems in which each decision is a choice from a discrete set of alternatives, uncertain chance events have discrete outcomes, and prior decisions may influence the probability distributions of uncertain chance events endogenously. In this paper, we develop the Decision Programming framework which extends the applicability of influence diagrams by developing mixed-integer linear programming formulations for solving such problems in the presence of many kinds of constraints. In particular, Decision Programming makes it possible to (i) solve problems in which earlier decisions cannot necessarily be recalled later, for instance, when decisions are taken by agents who cannot communicate with each other; (ii) accommodate a broad range of deterministic and chance constraints, including those based on resource consumption, logical dependencies or risk measures such as Conditional Value-at-Risk; and (iii) determine a...
This document can be stored and made available to the public on the open internet pages of Aalto ... more This document can be stored and made available to the public on the open internet pages of Aalto University. All other rights are reserved.
vehicles (BEVs) is studied. The studied problem involves a fleet of identical BEVs located at a c... more vehicles (BEVs) is studied. The studied problem involves a fleet of identical BEVs located at a central depot, a set of customers that must be serviced within given time windows, and a set of charging stations where the vehicles can recharge their batteries. The objective is to design a set of vehicle routes, each starting and ending at the depot, so that each customer is serviced exactly once and the total energy cost of the vehicle routes is minimized. Since BEVs have limited battery capacity and low recharging rate, charging station visits and recharging times must be explicitly considered in the route planning, wherefore most VRP variants are not sufficient in modeling the studied problem. Optimal routing of electric vehicles is not much studied in the optimization literature. The two most relevant models are the Green Vehicle Routing Problem (G-VRP) by Erdoğan & Miller-Hooks (2012) and the Electric VRP with Time Windows (E-VRPTW) by Schneider et al. (2013). The model considered in this thesis generalizes the E-VRPTW by also allowing the possibility of recharging a variable amount of energy at charging stations (variable recharging scheme) rather than performing a full recharge at every visit (fixed recharging scheme). This thesis introduces energy paths (e-paths) and proposes a new formulation of the studied problem based on non-dominated e-paths between every customer pair. The new formulation reduces the number of decision variables in the model and eliminates the need of imposing an artificial upper bound on the number of stops to a charging station, as is commonly done in previous models to keep their size acceptable. Some new preprocessing steps and valid inequalities are also presented to strengthen the LP-relaxation of the proposed formulation. Computational tests indicate that the new formulation provides considerable improvements over the standard formulation. Moreover, it is shown that significant reductions in the routing cost can be obtained in real-world routing problems by adopting the variable recharging scheme over the fixed one.
Influence diagrams are widely employed to represent multi-stage decision problems in which each d... more Influence diagrams are widely employed to represent multi-stage decision problems in which each decision is a choice from a discrete set of alternatives, uncertain chance events have discrete outcomes, and prior decisions may influence the probability distributions of uncertain chance events endogenously. In this paper, we develop the Decision Programming framework which extends the applicability of influence diagrams by developing mixedinteger linear programming formulations for solving such problems in the presence of many kinds of constraints. In particular, Decision Programming makes it possible to (i) solve problems in which earlier decisions cannot necessarily be recalled later, for instance, when decisions are taken by agents who cannot communicate with each other; (ii) accommodate a broad range of deterministic and chance constraints, including those based on resource consumption, logical dependencies or risk measures such as Conditional Value-at-Risk; and (iii) determine al...
Computers & Operations Research
Abstract We consider the Green Vehicle Routing Problem (G-VRP) which is an extension of the class... more Abstract We consider the Green Vehicle Routing Problem (G-VRP) which is an extension of the classical vehicle routing problem for alternative fuel vehicles. In the G-VRP, vehicles’ driving autonomy and possible refueling stops en-route are explicitly modeled. We propose a multi-start local search algorithm that consists of three phases. The first two phases iteratively construct new solutions, improve them by local search, and store all vehicle routes forming these solutions in a route pool. Phase three optimally combines vehicle routes in the route pool by solving a set partitioning problem and improves the final solution by local search. The algorithm is based on a multigraph reformulation of the G-VRP in which nodes correspond to customers and a depot, and arcs correspond to possible sequences of refueling stops for vehicles traveling between two nodes. All local search operators used by our algorithm are tailored to exploit this reformulation and do not explicitly deal with refueling stations. We report computational results on benchmark instances with up to ∼ 470 customers, showing that the algorithm is competitive with state-of-the-art heuristics.