The process of determining which projects to implement under a given budget, and which to defer until later, is central to the planning and management of highway systems. With a limited budget for construction, maintenance, and safety improvements, investments which will produce the optimal benefits must be chosen. This is often impossible to accomplish without the aid of a computer because of the complexity of the problem. Dynamic programming has been tested and verified as an efficient method for selecting priority projects to derive maximum benefits. The applicability of dynamic programming to the safety improvement program is demonstrated in this study.

There are several approaches to priority programming as it is related to the capital allocation problem. Benefit-cost, present worth, and rate-of-return calculations have traditionally been used as an integral part of the transportation planning process. Construction and maintenance programs continually face the task of having to assign priorities when insufficient funds are available to complete all projects. Safety improvement programs, which were initially funded through the Highway Safety Act of 1966 and expanded through the Federal-Aid Highway Act of 1973, have become so large that they are unmanageable without a clear and concise means of priority allocation.

A dynamic programming procedure was developed in this study which selects the optimal combination of safety improvement projects for a given budget. The type of dynamic programming being considered here is multistage. Multistage is defined as cost optimization of several projects, each with one or more alternatives. All safety improvement costs are dealt with in terms of present worth with consideration given to construction or installation cost, yearly maintenance cost, present interest rate, and the expected life of the improvement. The option of staging installation of safety improvements over a number of years was excluded from this analysis. All possible combinations of improvements were input as alternatives for each of the 61 projects involved in this study. The input consisted of the designated budget for the safety improvement program, the improvement cost, and the benefits derived from each improvement. The accuracy and reliability of dynamic programming is dependent upon the accuracy of benefits and costs used as input.

Report Date


Report Number

No. 412

Digital Object Identifier



Accepted for publication by the Transportation Research Board.