انتخاب شبکه در چارچوب مدیریت پرتفولیو برای برنامه ریزی حمل و نقل سازگارانه
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|21964||2011||14 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Transportation Research Part A: Policy and Practice, Volume 45, Issue 8, October 2011, Pages 765–778
A real option portfolio management framework is proposed to make use of an adaptive network design problem developed using stochastic dynamic programming methodologies. The framework is extended from Smit’s and Trigeorgis’ option portfolio framework to incorporate network synergies. The adaptive planning framework is defined and tested on a case study with time series origin–destination demand data. Historically, OD time series data is costly to obtain, and there has not been much need for it because most transportation models use a single time-invariant estimate based on deterministic forecasting of demand. Despite the high cost and institutional barriers of obtaining abundant OD time series data, we illustrate how having higher fidelity data along with an adaptive planning framework can result in a number of improved management strategies. An insertion heuristic is adopted to run the lower bound adaptive network design problem for a coarse Iran network with 834 nodes, 1121 links, and 10 years of time series data for 71,795 OD pairs.
Conventional practice in transportation planning relies on a passive approach to project investment. Each project is typically evaluated in a single long range future forecast year, and then compared against other projects on a year-to-year basis for funding without any adaptation to changing conditions over time. Although project investment decisions are made on a year-to-year basis with the long term plan in mind (Kim et al., 2008 and Salling and Banister, 2009), there is a lack of a systematic framework to continually evaluate projects against each other under changing conditions. There is increasing interest from public sector agencies to adopt a more active management style of project management. In an ongoing National Cooperative Highway Research Program (NCHRP) study, Caplice and Dahlburg (2011) point to the flaws of forecasting with point estimates and suggest a qualitative scenario planning approach to identify long term possible future scenarios which can be monitored using news updates as “sensors in the ground”. Nonetheless, recent surveys of public agencies continue to report an ongoing mismatch between the state of practice and the state of the art and the desire to address that gap (Hatzopoulou and Miller, 2009). Given the gap, one might expect to find a robust academic literature devoted to time dependent transportation planning. Multi-period transportation network design was proposed as early as the 1970s (Steenbrink, 1974). However, a number of important advances in this field were only made in recent years due to modeling and computational complexities. Wei and Schonfeld (1993) and Kim et al. (2008) provide heuristics for solving multi-period discrete network design problems under a deterministic setting. Szeto and Lo published a number of papers in the area of deterministic multi-period continuous network design, the most recent of which include tolling (Szeto and Lo, 2008) and cost recovery (Lo and Szeto, 2009). However, practitioners’ skepticism of state-of-the-art deterministic network design models (Hatzopoulou and Miller, 2009) call to question whether purely deterministic approaches are sufficient under a world of greater uncertainty. Sensitivity approaches have been developed to address this concern for deterministic single period model solutions (Salling and Banister, 2009) as well as for multi-period network design (Szeto and Lo, 2005). The conclusions of these studies indicate a greater need for more flexible, adaptive planning approaches. A number of attempts at incorporating uncertainty in network design and investment models have been considered, although stochastic demand and capacity are generally dealt with using stationary, static distributions with scenario planning approaches. Stochastic programming efforts have been applied to bi-level network design problems where congestion effects occur (Waller and Ziliaskopoulos, 2001 and Chen and Yang, 2004) as well as in facility location and inventory management problems (Shu et al., 2005 and Snyder et al., 2007). These studies do not consider multi-period, time-dependent design decisions with stochastic variables. Ukkusuri and Patil (2009) propose a multi-period stochastic network design problem formulation that accounts for elastic demand. They formulate the model of allocating design variables to a number of links as a mathematical program with equilibrium constraints (MPEC) with stochastic demand over multiple time periods. However, this formulation does not treat future period investment decisions as explicit options that depend on the realization of all the stochastic elements up to that point, as an adapted process. A truly flexible planning approach requires consideration of multiple periods and adaptive decision-making which is essentially stochastic dynamic programming. Chow and Regan (2011b) propose a Link Investment Deferral Option Set (LIDOS) and a lower bound solution for this problem. LIDOS is a stochastic dynamic programming approach to network design, which is re-christened as an adaptive network design problem (ANDP) for simplicity. In the problem, each link investment is treated as a separate option whose investment could impact the value of other link investments. Details of the ANDP can be found in Chow and Regan (2011b). The current research serves as a companion piece; its contribution is the definition of a transportation planning framework that embodies the ANDP, using concepts from portfolio management that incorporates real option strategies. Both concepts of portfolio management and real options are defined in Section 2. The framework is empirically tested on a network where time series OD data is available so that illustrative comparisons of planning strategies can be made against more conventional and current state of the art practices.
نتیجه گیری انگلیسی
6.1. Policy implications A number of lessons can be learned from this research study. First, there is an agreement in the literature that static planning is insufficient in a world of greater uncertainty, but up until now it has been restricted due to lack of data and lack of analytical tools. Given the recent development of an adaptive network design approach, an adaptive portfolio management framework is appealing to transportation network managers. Such a framework can be implemented if data is available using the ANDP as an underlying engine for project evaluation and decision support. While many of the assumptions simplify the case study and prevent actual policy-making decisions from being recommended, our model does make use of real OD demand and a coarse intercity road network, providing an example of how this data can be exploited. It is clear from the results and observations that even when treating only truck ODs as sources of volatility there is a significant amount of leverage in employing an adaptive planning approach. Even with a limited 5 year planning horizon, the adaptive planning framework is able to weather the volatile demand environment observed during those 5 years. Applying the framework to transportation planning allows decision-makers to dynamically manage their projects from a central authority and compare immediate benefits to various risks. The framework allows managers to continuously monitor their projects and provides visual flags for attention with the option space. Fig. 5 and Section 5.3.1 clearly show that even a continuous set of deferral decisions over a 5 year period can provide a manager with an abundance of information for planning strategies under an option portfolio framework and option space. This is a call to arms to planning agencies – MPOs and transportation agencies – to consider adopting a portfolio management planning framework and acquiring the data for it. Due to the general applicability of the framework to different network evaluation methods, it can be adopted by both public sector planning organizations and private firms. Case studies examining supply chains of global technology giants like HP or Amazon.com can illustrate the benefit of evaluating network improvement projects that incorporate some of the effects of network synergies using the adaptive network design lower bound solution. 6.2. Future research One area that is heavily discussed in Smit and Trigeorgis (2006), but not touched upon here, is the trade-off between deferral and commitment. Deferring an investment has the benefit of gathering new information. On the other hand, from a game theoretic perspective, committing to an investment sends signals to competitors which can force them to act in ways that may be strategically beneficial to the organization. This is the idea behind first mover advantages, and Smit and Trigeorgis’ framework incorporates this game theoretic approach. When other players are ignored there tends to be a bias towards deferring investments. Taking account of the multiple players in a network setting would allow decision-makers to truly evaluate the benefits of network improvement timing against agency–agency or agency–public interactions. Such a framework provides a more balanced and realistic view of network strategic planning that warrants further study. Examples of benefits of a game theoretic network portfolio framework include high speed rail investment in an airline dominated industry; cooperative bidding of projects between state transportation departments and local city agencies; and investment of alternative fuel infrastructure by competing providers when consumer demand is responsive to first movers. Future research should also consider other heuristics such as tabu search or genetic algorithm to allow a greater number of projects for evaluation; large-dimensional covariance matrix estimation for considering correlated OD stochastic processes; and combining this model with integrated land use models so that infrastructure investment decision-making can be dynamically integrated along with activity interactions and travel demand. Further, this problem lends itself to distributed programming because the sequential lengthy evaluations can each be viewed as independent “copies” of the problem.