Virtual Transportation Manager : A web-based system for transportation optimization in a network of business units

Transcription

1 Virtual Transportation Manager : A web-based system for transportation optimization in a network of business units Jean-François Audy 1, Sophie D Amours 2, Louis-Martin Rousseau 3, Jean Favreau 4 and Philippe Marier 5 1 Doctoral student, Department of Mechanical Engineering, CIRRELT, FORAC Research Consortium, Université Laval, jean-francois.audy@cirrelt.ca 2 Professor, Department of Mechanical Engineering, CIRRELT, FORAC Research Consortium, Université Laval, sophie.damours@forac.ulaval.ca 3 Professor, Department of Mathematics and Industrial Engineering, CIRRELT, École Polytechnique de Montréal, louis-martin.rousseau@polymtl.ca 4 Group leader - Precision Forestry and Logistics, FPInnovations-FERIC, jean-f@mtl.feric.ca 5 Research professional, FORAC Research Consortium, Université Laval, philippe.marier@forac.ulaval.ca Summary In this paper, the decision support system VTM, for Virtual Transportation Manager, is presented. The VTM is a web-based system developed to allow collaborative route planning among many transportation managers, from different companies or business units forming a coalition. VTM is supporting the coalition s administrator, dispatcher as well as its operating members (e.g. forest companies). The role of each of these three VTM users as well as their responsibilities within the system is detailed in the paper. Also, through the description of a complete transportation planning sequence on the VTM, the different modules of the system and their functionalities are introduced. With the support of Groupe Transforêt, the VTM was implemented to simulate its utilization over almost one year. Numerical results showing significant savings in terms of cost, fuel consumption and traveling distance are reported. Introduction Currently, carriers play an important role in the day to day operations of Canadian forest companies and assist companies in attaining their business objectives. Many economic considerations (e.g. reduction of inventory costs, additional processing costs generated by poor fiber freshness, rising fuel prices, worldwide competition) contribute to increase the importance of transportation logistics in the wood supply chain. In the Canadian industry, the standard haulage route is a series of direct trips where the truck travels loaded from the origin (i.e. harvest area or transit and stocking yard) to the delivery site (i.e. mill or transit and stocking wood yard) and travels back empty. This results in poor efficiency with half of the hauling distance empty, and even more, if we consider the distance from/to a truck s depot before and after the route. In the Canadian province of Quebec, transportation and handling operations represent an average of 21,7% and 20,3% of the wood cubic meter cost at the mill in the softwood and mixedwood-hardwood forests respectively (Consultants forestiers DGR inc., 2003). On average, these operations represent an expense of up to half a million Canadian dollars a year for forest companies. Making these operations more efficient represents an excellent cost-saving opportunity for forest companies. One of the main explanations for these inefficiencies is the decentralized organization of transportation planning, which is in the hands of local transportation managers. Even if a local manager wants to improve

2 his transportation efficiency, the potential is usually quite low within the wood flow pattern of his local transportation needs. However, in many forest regions, pooling the transportation needs of the managers within the region leads to a wood flow pattern with a greater potential for transportation efficiency improvement. Essentially, the greater potential with the coordination of the wood flow among many managers from different companies or divisions within a company, comes from three opportunities: wood bartering, backhauling and better routing. In wood bartering, volumes of some supply points are exchanged between the companies or divisions to reduce the total travel distance. Backhauling and better routing is used to find enhanced routes that combine volume of different companies or divisions to avoid direct trips. There exists a number of decision support systems in wood transportation addressing one or many of these opportunities, see Åkar-web (Eriksson and Rönnqvist, 2003), ASICAM (Weintraub et al., 1996), EPO2 (Linnainmaa et al., 1995), FlowOpt (Forsberg et al. 2005), RuttOpt (Andersson et al., 2007) and MaxTour (Gingras et al. 2005). Potential cost-savings in the range of 3-30% are reported. In this paper, a new DSS in wood transportation is presented; the VTM stands for Virtual Transportation Manager. The main difference between VTM and the previous DSS is its collaborative architecture. The development of the VTM is a partnership project between FPInnovations-FERIC and FORAC Research Consortium spanning from 2003 to The following section of this paper details the VTM system. Then, a case study is introduced and the results of a ten months simulation are presented. Finally, we provide some concluding remarks and considerations for future research. The VTM System The VTM is a web-based system using information technologies to consolidate and manage transportation data from several business units in order to improve the efficiency of their routing. The VTM allows the generation of a set of enhanced routes and its management. There are three roles in a VTM: administrator, dispatcher and member. Each role has different responsibilities and access rights to the several modules of the VTM as well as different ways to access companies data. The administrator role is assigned to the manager of the system: he has access to all modules and all data of each company but is essentially responsible for membership management. Typically, the VTM system would be hosted by a third party. Other types of hosting are possible, such as the hosting of the system by one of the members of the VTM, but this could raise the issue of data confidentiality. The dispatcher role is assigned to the central management of the transportation operations on the VTM: he is entitled to the same access rights as the administrator, except for the module on membership management. The dispatcher is responsible for the planning engine that generates the routes and the transportation data accuracy and update. The member role is assigned to the forest companies. Thus, to access a VTM, a company would need to become a member of that VTM. Once a company is a member, at least one user profile needs to be created for that member. By its personalized secured profile, the users perform operations on the VTM for the member company. It is this member-user(s) relation that allows complete visibility of the data belonging to a company for all its users and, conversely, to guarantee companies private data confidentiality between them. Each company has access to the modules related to their transportation requests (i.e. transportation requests data import, site location confirmation and inventory mapping query). A transportation request is defined by a product type and attribute(s), an origin and a destination site, a volume, a weight and an earliest pick-up date at the origin and a latest delivery date at the destination. In contrast to most general freight transportation, in wood transportation it is not required to entirely pick up a transportation request (i.e. splits are allowed). Usually, there is one administrator, one dispatcher and many members in a VTM. A VTM with only one member is defined as an intra-organizational VTM, meaning that all users of the VTM belong to the same company. An intra-organizational VTM has been implemented in the case study used below. A VTM with at least two members is defined as an inter-organizational VTM, meaning that users of the VTM come

3 from different companies. It is mainly in an inter-organizational VTM that all the system-customized features on the data confidentiality issue are essential to support a sustainable collaboration. The following subsections detail the transportation planning sequence in a VTM. Users import of their transportation requests The VTM users put their transportation requests on the VTM using an import module based on the xml file format. As it was done in the case study, it is possible to develop an automatic Internet import service transferring the data from the members inventory systems to the VTM. At each new import, an update process is performed by the VTM, comparing the transportation requests already on the VTM to the transportation requests update. Thus, new transportation requests in the import are added to the VTM, transportation requests already on the VTM but not present in the import are deleted, and finally, transportation requests already on the VTM with new parameters value in the import are modified. Once a user has imported its transportation requests on the VTM, he can consult the different data related to these transportation requests. The most interesting module for a user is the inventory mapping query module. Figure 1 shows a screenshot of the inventory query module and its mapping result. The arrows show the flows of request to ship or to receive and the circled sizes represent the weight or the volume of the requests at an origin or a destination site. Just being able to see on a map the requests that need to be transported provided an improvement in transportation efficiency in the case study. Figure 1: Screenshots from VTM on the inventory query module (left) and its mapping result (right) The VTM members have to agree on certain naming conventions in the definition of their transportation requests. These conventions are required for the system to be able to perform data consolidation in several modules. These conventions are: - A product type and attributes convention: VTM has been built to support the planning of different types of products (e.g. logs, chips, lumber) with different attributes (e.g. firm-spruce chips, pine chips) to catch the potential additional savings provided by multi-use trailer, as reported by Brown et al. (2003) and Gingras et al. (2005) in their respective case studies on multi-use trailers for logs and bulk fiber transportation (i.e. chips, sawdust, bark and residue). Adhering to this convention is essential in transport planning in which compatibility and incompatibility relations must be respected. For example, a compatibility matrix between truck types and product types specifies which product type can go on a given truck type and an incompatibility relations product attribute/product attribute specifies that some products with given different attributes cannot be hauled at the same time. Finally, this convention specifies the volume and weight units of the different products on the VTM that need to be followed by the members when importing their transportation needs. - A sharing site convention: there are two kinds of status for a site: private or shared. This status is attributed by the first user who, by importing its transportation request, imports a new origin or

4 destination site on the VTM. After a confirmation of the new site location on an electronic map in the site location confirmation module, the user must decide if the site and its potential handling resource will be viewed by all the users (shared) or only by the users who belong to the same company (private). Usually, most sites are shared. Private sites are mostly for business considerations such as competition between two or more members for a supplier or a customer. As for the previous convention, respect for this convention is essential in the transport planning for some incompatibility relations of large truck type / narrow site or unauthorized truck type / site located in a particular legislative territory. Dispatcher sets up the transportation data Before he starts the planning, the dispatcher must perform tasks related to transportation data accuracy and update. - set up the shared and private sites information related to transport (i.e. site access time windows and on-site handling resource time windows). When the site is private, this information must be sent to the dispatcher by the company the private site belongs to. If a new shared site has been added by a user, the dispatcher must verify that the same site with possibly a different name does not already exist. The VTM automatic calculation process of the traveling time and distance matrix between the new site and all the other sites already on the VTM must also be verified by the dispatcher to avoid potential zero value if a new site is not connected by the system on the road network. In the case study used below, the company does not have a GIS to support the time/distance calculation and an electronic map creation process is necessary in some modules. However, all its sites being directly connected to the public road network, Microsoft MapPoint was utilized in both processes. - set up the trucking capacities. Since the VTM has not been built to manage a private or dedicated fleet of trucks (and their carriers) spread over a set of specific depot sites, the trucking capacities are represented on the basis of transport zone. A transport zone includes one or many of the private and shared sites. In each zone, a capacity, in number of trucks of the same type (i.e. truck with similar transportation relevant characteristics such as maximum weight and volume, with or without crane, fuel consumption, etc), is set by the dispatcher. A calendar of availability, usually 5 or 7 days a week, could be added to this capacity. The zones creation as well as the capacity is based on discussion between the dispatcher and the different members. The dispatcher aims to represent as realistically as possible the available truck capacity obtained through the members agreements with carriers. Also, in each zone, the trucks of the zone are linked to a pseudo-depot site, usually located in the middle of the zone. Payment methods used in practice by the industry fix the transportation price on the basis that a route starts and finishes at the first pickup site. This practice is applied in the route planning except when both the first pickup site and the last delivery site on the route are outside the zone of a truck. In this situation, the route starts and finishes at the 'pseudo-depot' to reflect the additional costs in empty traveling distances from and back to the zone. - set up the truck costs value. VTM uses a cost function representative of the operational cost of a truck, i.e. without the variable profit (or loss) percentage included in the rate tables of a carrier. The cost function includes an hourly working cost for each type of truck, a fuel cost based on three engine fuel consumption functions (i.e. idling, handling for a truck with a crane and carrying, which depends on the total weight of the truck) and, finally, a fixed cost for an overnight route (driver compensation). - set up the driver regulation and other rules. Usually, there are three basic rules in transportation legislation that regulate working and driving hours for drivers of commercial carriers: (1) a maximum driving time allowed per working shift; (2) a maximum working time allowed per working shift and (3) a minimum consecutive rest time for the driver each day. The loading and unloading times being long, in practice, route generation is restricted by the second rule rather than the first one. Thus, the possible set up of (2) and (3) has been integrated in the VTM as well as two other limits related to general practices and industry standards. They are: (4) a route must start during a normal working

5 day and, to reduce the number of nights away from home for drivers, (5) a maximum number of consecutive working shifts on a route. Dispatcher starts the planning of the transportation requests Once all transportation data are ready for planning, the dispatcher accesses the planning engine module. After a setting of the planning parameters (e.g. time horizon, planning start date, fuel price, loading (slow) and unloading (fast) unit time for truck type with (fast) and without (slow) crane), this module allows the starting of an optimization software link to the VTM. Using OPL Studio 3.7 from ILOG software, a greedy & repetitive solution methodology imbedding heuristics, a taboo list and constraint programming models have been developed. The choice of constraint programming is explained by its two major features reported by Gendreau (2002): expressivity for problem complexities descriptions and flexibility for problem resolution possibilities. The main idea behind the solution methodology is to use optimized sequences of origin and destination sites, called itinerary, as a support to build the routes. Two sets of itineraries are used by the solution methodology: a pre-optimized itinerary set, obtained with the MaxTour tactical level decision support system of Gingras (2005) as in the case study used below, and an itinerary set created during the solution methodology. The general framework of the solution methodology is summarized in Marier et al. (2007). During the implementation of the VTM for the case study used below, some additional development of the solution methodology has been necessary. For example, before the planning, a customized process had to be developed to consolidate in transportation requests all the logs stored individually in the inventory system of the case study. Also, a site aggregation/disaggregation procedure before/after solving the MaxTour tool has been developed to obtain a pre-optimized itinerary set adapted to the many less-thantruckload transportation requests of the case study. Moreover, a short fixed time before/after pickup and/or delivery on a site has been added to the unit loaded/unloaded time to avoid undervaluation of the duration required to perform the pickup/delivery operation of a less-than-truckload transportation request. Dispatcher manages the planned routes Once the planning is done, the dispatcher is responsible for the management of the planned routes. Since this management could be done in several ways depending on the members of the VTM, a simple but useful module that supports the dispatcher in a manual management of the route has been developed. Figure 2 shows the route management module in which the dispatcher could, one route at a time, analyze statistics and key indicators of the route as well as all the pickup/delivery operations on each segment of the route. Figure 2: Screenshots from VTM on the route management module giving the route statistics (left) and the detailed route segments (right)

6 Using the carriers contacts provided by the member of the VTM, the dispatcher must find a carrier on each route and then ensure the follow-up on the completion of the route by this carrier. In the module showing the planned routes, three kinds of route status are used: planned, approved and completed. The planned status is the default status of a route. If the route status is not modified before the next planning, the route is deleted to allow a new routing possibility to the transportation requests on the route. The next status, approved, is attributed to a route when the dispatcher has proposed the route to a carrier or if a carrier has accepted the route (reservation). A route with this status will never be deleted when a new planning is performed. If the carrier refuses the proposed route or cancels its reservation, it s possible to change the status. Finally, the status completed is attributed to a route that was carried out by a carrier. Case study: Groupe Transforêt The case study involves a high-value hardwood log-specialized supplier operating in eastern Canada, Groupe Transforêt, which buys logs from a network of more than 4500 wood producer/supplier-mills to resell them in its network of customer-mills. This large territory is separated into regions, each being the responsibility of one coordinator (a manager) who is in charge of the purchasing and transportation planning of the logs purchased inside the region. Figure 3 illustrates the coordinator region and business network of GT. Figure 3 : Coordinator regions and business network of Groupe Transforêt Each coordinator uses a hand-held computer to directly register all purchased logs in the field. An inhouse Internet based inventory system is used to transfer invoicing information on the purchase from all regional coordinators to GT s central administration office. Even if GT is only a single company, the transportation planning is organized as if each regional coordinator were an independent company. The transportation requests of GT are almost all less-than-truckload and few represent a truckload or more. Also, transportation requests are characterized by long one-way delivery distances, from 50 km up to 700 km, compared to the average one way delivery distance of 140 km in Canadian softwood operations (Hillman and Michaelsen, 2006). To carry-out its transportation planning operations, GT coordinators utilize several truck types owned by several carriers. In almost all regions, five truck types are available with 13.5, 31.5, 34.5, 37.5 and 41 ton truck trailer capacities. The two smaller truck types have cranes and they are essential for pickup at some origins without on-site handling equipment. For a complete description of the operations at Groupe Transforêt as well as the routing problem, see Marier et al. (2007). The profitability of GT can be roughly summarized by the difference between the resale price of a log and its purchase price minus the operation costs to deliver the log to the customer-mill. Transportation costs account for a significant proportion of these operation costs. With the knowledge that there are transportation flows between several regions, collaborative transportation planning could lead to more

7 efficient routes which are not possible in the current governing mode. This increase in transport efficiency would lead to a reduction in transportation costs which in turn, would lead to an increase in profitability for GT. Numerical tests and results In several case studies on collaborative planning, the savings are defined as the difference between the cost of the optimized collaborative plan (i.e. all the local managers of the divisions or companies together) compared with the sum of the cost of each optimized individual plan (i.e. each local manager of the division or company alone). We follow this approach to set the savings in the following numerical tests. Numerical tests for almost one year (July 2006 to April 2007) of GT s transportation requests were realized. Every two weeks, each coordinator's individual planning was done and compared to the results of a collaborative planning approach. Period Resolution time (second) Table 1: Numerical results Reduction Cost ($) Fuel (L) Distance (Km) 1 543,6 ± 233,5 9,8% ± 4,7% 7,1% ± 3,5% 13,0% ± 5,6% 2 647,3 ± 234,0 7,3% ± 3,8% 4,3% ± 4,6% 11,1% ± 4,6% 3 620,4 ± 149,2 10,5% ± 4,2% 8,3% ± 3,1% 12,1% ± 4,3% Table 1 presents the average and standard deviation results according to three distinct operating periods for GT for a total of 5, 7 and 9 two weeks planning instances respectively in each operating period. For each period, we provide the average resolution time and percentage of reduction in cost, fuel consumption and traveling distance. Average cost-saving opportunities in the range of 7,3% to 10,5% exist in the collaboration of the six coordinators. Conclusion In this paper, we presented the collaborative decision support system VTM. Then, we introduced the Groupe Transforêt case study and presented the results of numerical tests. These showed that collaborative transportation planning through the VTM system allows significant savings in terms of cost, fuel consumption and traveling distance. In order to evaluate opportunities to generate more cost-saving through collaboration on the VTM system, larger case studies networks must be tested in future work. As well, economic evaluation on the use of multi-use trailers must be conducted to evaluate the additional benefits of opening these networks to other wood products. Implementation challenges such as benefit/loss sharing of the VTM, carrier selection and competition, membership and dispatcher/system expenses, route billing, etc. remain to be addressed. Acknowledgements The authors wish to thank other members of the VTM development team from FORAC and FPInnovations-FERIC as well as Groupe Transforêt. References Andersson, G., Flisberg, P. and M. Rönnqvist RuttOpt A decision support system for routing of logging trucks, Scandinavian Working Papers in Economics, NHH Discussion Paper 16 Brown, M., Michaelsen, J., Hickman, A. and Y. Provencher Potential for multi-use trailers in the Canadian forest industry. CR , Forest Engineering Research Institute of Canada, Canada Consultants forestiers DGR inc Wood supply cost in province of Quebec Results of 2003 survey for Fir-Spruce-Pine-Tamarack species. Consultants forestiers DGR inc, Canada. [In French]

8 Eriksson, J. and M. Ronnqvist Transportation and route planning: Åkar-web - a web-based planning system. In the Proceedings of: 2 nd Forest Engineering Conference, Växjö, Sweden, May 12-15, pp Frisk, M., Jörnsten, K., Göthe-Lundgren, M. and M. Rönnqvist Cost allocation in collaborative forest transportation, Scandinavian Working Papers in Economics, NHH Discussion Paper 15. Forsberg, M., Frisk, M. and M. Rönnqvist FlowOpt, a decision support tool for strategic and tactical transportation planning in forestry. International Journal of Forest Engineering 16(2): Gendreau, M Constraint programming and operations research: comments from an operations researcher. Journal of Heuristics 8: Gingras, C Transport optimization in the forest industry. Master s thesis No. 82, École des Hautes Études Commerciales de Montréal, Canada. [In French] Hillman, D. and J. Michaelsen Competitiveness of on-road transport for the Manitoba forest industry. Contract Report CR , FERIC, Canada. Linnainmaa, S., Savola, J. and O. Jokinen EPO: A knowledge based system for wood procurement management. In the Proceedings of: 7 th Innovative applications of artificial intelligence conference, Montreal, Canada, August 21-23, pp Marier, P., Audy, J.-F., Gingras, C. and S. D Amours Collaborative wood transportation with the Virtual Transportation Manager. International Scientific Conference on Hardwood Processing, September 24-26, Quebec City, Canada, 10 p. [accepted] Weintraub, A., Epstein, R., Morales, R., Seron, J. and P. Traverso A truck scheduling system improves efficiency in the forest industries. Interfaces 26(4):1-12.