Computer-Aided Civil and Infrastructure Engineering 21 (2006) 193–204 True Optimization of Pavement Maintenance Options with What-If Models Koji Tsunokawa, ∗ Dinh Van Hiep & Riaz Ul-Islam Graduate School of Civil and Environmental Engineering, Saitama University, 255 Shimo-Okubo, Sakura-ku, Saitama-shi, Saitama-ken, 338-8570, Japan Abstract: A prototypical problem road agencies are faced with is to ﬁnd the optimal application schedule of maintenance works for a given road section. To solve such problems what-if models such as the road trans- port investment model (RTIM), the highway economic requirements system (HERS), and the highway develop- ment and management tool (HDM-4) are widely used to predict the consequences of different maintenance op- tions. With these models maintenance options to be com- pared must be exogenously speciﬁed by an analyst, and the “optimization” with these routines simply chooses the best among those compared. As there are usually inﬁnite numbers of options, it is impossible to exhaust all of them and only suboptimal optimizers may be found with this approach. The present article proposes the use of gradient search methods with what-if models to ﬁnd the true op- tima without requiring exogenously speciﬁed alternatives. It demonstrates through a case study the feasibility of the use of the steepest descent method and the conjugate gra- dient method along with HDM-4 to ﬁnd the true optimum maintenance options. 1 INTRODUCTION 1.1 Background Road managing agencies strive to maintain their net- works by selecting maintenance alternatives that mini- mize total transport cost (TTC) or maximize beneﬁts to the stakeholders in the long term. This requires the use of analysis systems that can predict the costs and bene- ∗ To whom correspondence should be addressed. E-mail: koji-t@post. saitama-u.ac.jp. ﬁts of different maintenance options over time. Accord- ingly, what-if models of pavement management analysis such as RTIM (Cundill, 1993), HERS (FHWA, 1998), and the highway development and management tool (HDM-4) (PIARC, 2002) that predict the consequences of different maintenance options are often employed to “optimize” a maintenance schedule. These models are built on sets of relationships for predicting pave- ment performance and corresponding user beneﬁts over time taking into account the speciﬁed maintenance op- tions. The relationships in the what-if models are the product of extensive research undertaken during the last four decades (see, e.g., Moavenzadeh et al., 1971; Abaynayaka et al., 1977; CRRI, 1982; GEIPOT, 1982; Hide, 1982; Watanatada et al., 1987; NDLI, 1995; Ben- nett, 1996). The major weakness of the what-if models is that they require exogenously speciﬁed maintenance options to be compared. Therefore, when they are used to “optimize” maintenance options, which is often the case, they are not optimizing them in its true sense; they are merely used to ﬁnd the best among those exogenously speciﬁed. As there are inﬁnite numbers of options, it is impossi- ble to exhaust all of them to ﬁnd the true optimizer and only suboptimal options may be found. Usually, alterna- tives to be compared are generated with “engineering judgment,” and the situation may not be so bad as a mere “trial and error” (see, e.g., Bhandari et al., 1987; Riley et al., 1994; Tsunokawa et al., 2002). However, it does not diminish the necessity of developing a sys- tematic and efﬁcient search procedure that obtains true, global optimizers while obviating exogenously speciﬁed alternatives. This article shows how standard optimization algo- rithms can be used with what-if models in a systematic way to ﬁnd global optimum maintenance options. The C  2006 Computer-Aided Civil and Infrastructure Engineering. Published by Blackwell Publishing, 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington Road, Oxford OX4 2DQ, UK.