Karthik K Srinivasan

Sensitivity analyses are important parts of both studying complex systems and measuring the variation in input parameters on the response. They are useful to decision makers for understanding the robustness of the optimal solution that they are to adapt to variations of the parameters of the problem. The sensitivity of the optimal solution of a project-level pavement management problem is analyzed, and the robustness of the optimal solution to the interventions and the timing, cost, and benefit are investigated. The input parameters, which affect the optimal maintenance solution, are identified as the structural and functional condition parameters (defined in terms of deflection and roughness, respectively, at the beginning of the analysis period), traffic volume, growth rate, and discount rate. The problem of computing the optimal treatment and timing for a given budget level is modeled as a mixed integer nonlinear optimization problem and solved by using a computationally efficient network-optimization technique. The benefits are evaluated by considering the pavement performance and are quantified as the area between the performance curve and the threshold values. The optimal budget required for pavements in different structural and functional conditions as well as traffic levels is presented. The effect of initial pavement condition on the optimal maintenance actions as well as their timings is studied. The result of the sensitivity analysis showed that the cumulative standard axle loads and traffic growth rate have a significant effect on the selection and timing of rehabilitation and preventive maintenance actions. The effect of the discount rate on the maintenance management decisions is also presented.

This study analyzes changes in sociodemographic, activity, land use, and mobility patterns and their effects on travel dimensions in the context of a developing country. More specifically, increase in vehicle ownership (both two-wheelers and cars) and changes in mode choice over time are observed and analyzed with the use of household data from Chennai, India. Three sources of dynamics are analyzed: exogenous variable dynamics, sensitivity changes over time, and the influence of lagged and persistent effects. The key drivers of growth in travel demand include the increase in vehicle ownership, the number of workers, and the increase in female drivers. The influence of social and technological factors on vehicle ownership and mode choice such as peer pressure and mobile phone ownership are also significant. In addition, the effect of land use, accessibility, and activity has been investigated. Results show significant evidence of differences in travel decisions across different user segments (on the basis of driving knowledge and vehicle-worker ratio) and over time. The proposed disaggregate models provide a reasonably good description (goodness of fit is 47% to 64%) of the observed changes in travel patterns. The findings and results assume importance in the context of increasing congestion, declining public transportation share, and the imminent need for enhancing urban transportation system capacity in cities of developing countries.

This paper investigates the day-to-day dynamics in an urban traffic network induced by joint route and departure time switching dynamics in commuter decisions under real-time information. The role of switching inertia, the user's sensitivity to experienced and lagged system attributes, and travel demand management measures on day-to-day evolution are analyzed. A simulation-based day-to-day network analysis framework is developed by incorporating empirically calibrated user behavior models under information. This model can be used to analyze network reliability and stability measures. Because it can model users' responses to information and experience, this framework permits relaxing the assumption that user behavior remains unchanged. Therefore, it can be used to model the transient effects due to various types of system perturbations and shocks. Computational experiments are used to investigate the effect of the aforementioned factors. The findings suggest that trip time variability and lateness propensity increase under high congestion. Under joint switching dynamics, a significant deviation from equilibrium is observed even after nearly 2 months. Departure time switching behavior appears to exert a greater influence than route switching on day-to-day dynamics, and transportation control measures can improve trip time and network reliability. These results have important implications for congestion mitigation strategies, network reliability assessment, and evaluation of intelligent transportation system technologies.

This paper investigates mode choice decisions of workers in Chennai City, a metropolis in India. Mode choice in developing countries such as India may differ significantly from their developed counterparts in several respects. These differences pertain to vehicle types (two-wheelers versus four-wheelers), vehicle ownership levels and growth, wide variation in socioeconomic characteristics, perception of subjective factors, and variability in choice set, among other factors. Toward understanding these differences, this paper investigates the role of the following factors on mode choice: (a) differences between mode choice propensity for two-wheelers and four-wheelers; (b) differences in sensitivity to travel time and cost across different user groups based on captivity effects [captive (no vehicles), semicaptive (fewer vehicles than workers), and choice segments]; (c) alternative means of representing the unavailability or infeasibility of some alternatives to some users; and (d) the influence of subjective factors. To achieve these objectives, a series of disaggregate mode choice models are developed to capture the aforementioned effects, on the basis of data from 550 workers in Chennai City. Six alternatives are considered: two-wheeler, car, bus, train, nonmotorized, and other modes. The empirical results underscore significant variations in two-wheeler and car choice propensities. Furthermore, significant differences in sensitivity to travel time, cost, and vehicle availability are observed across different user segments. The cost sensitivity to various modes reduces as the commute distance increases. Results indicate that disregarding the aforementioned effects can lead to poor model fit, biased coefficients, and erroneous forecasts. These findings have important implications for the evaluation of transit ridership improvement strategies and demand forecasting in developing countries.

With the advent of more flexible discrete choice models, the analysis of heterogeneity at the observed and unobserved levels is receiving increasing attention. However, heterogeneity in decision rules has hardly been investigated in the mode choice context. This study proposes a heterogeneous decision rule model of mode choice incorporating utility maximization and disutility minimization using empirical data from Chennai City, India. The two decision rules may yield different estimates of mode choice probabilities if the error structure is not symmetric. Therefore, a heterogeneous decision rule model is estimated by postulating separate choice behaviors for each decision segment. Because the decision rule remains latent, individuals are probabilistically assigned to the two segments. The membership propensity of belonging to each class is modeled by using a binary logit form. The performance of the proposed heterogeneous decision rule (HDR) model is compared with the pure utility maximization, pure disutility minimization, and heterogeneous latent class models. The results reveal that the HDR model outperforms these alternative specifications. Further, significant differences are observed across the decision segments for aggregate modal shares, intrinsic preference for different modes, sensitivity to modal attributes, role of subjective factors, and the effect of activity patterns and accessibility. Factors influencing the decision-segment membership propensity are also identified. These findings have important behavioral and practical implications for analysis and evaluation of travel demand management measures aimed at sustainable urban transportation systems, congestion mitigation, and transit improvement.

The rapid and continuing changes in travel and mobility needs in India over the last decade necessitates the development and use of dynamic models for travel demand forecasting rather than cross-sectional models. In this context, this paper investigates mode choice dynamics among workers in Chennai city, India over a period of five years (1999-2004). Dynamics in mode choice is captured at four levels: exogenous variable change, state-dependence, changes in users' sensitivity to attributes, and unobserved error terms. The results show that the dynamic models provide a substantial improvement (of over 500 log-likelihood points and ρ2 increases from 44% to 68%) over the cross-sectional model. The performance was compared using two illustrative policy scenarios with important methodological and practical implications. The results indicate that cross-sectional models tend to provide inflated estimates of potential improvement measures. Improving the Level of Service (LOS) alone will not produce the anticipated benefits to transit agencies, as it fails to overcome the persistent inertia captured in the state-dependence factors. The results and models have important applications in the context of growing motorization and congestion management in developing countries. © Springer Science+Business Media, LLC 2007.

This paper focuses on the application of the kernel logit formulation to model dynamic discrete choice data. A dynamic kernel logit (DKL) formulation with normal errors is presented to model unordered discrete choice panel data. Investigating the theoretical foundations of the kernel logit model, it is demonstrated that the mixed logit error structure converges in distribution asymptotically to a suitable multivariate normal error structure. This result provides support for both cross-sectional kernel logit (CKL) and DKL models with normal errors. The calibration, identification, and specification issues associated with the latter model are also discussed. The performance of the proposed DKL model is assessed from the perspective of computational efficiency and estimate accuracy relative to the multinomial probit (MNP) model using a series of numerical experiments. Complexity analysis reveals that the DKL has a lower computational complexity than the MNP frequency simulator, which has an exponential complexity. Thus, for choice situations with a large number of alternatives (J) in each time period, and/or large number of time periods (T), the DKL model is faster than the corresponding MNP by more than an order of magnitude. This is also confirmed by computational experiments conducted using 32 synthetic data sets. The computational performance of the DKL relative to MNP appears to be the result of a trade-off between the number of Monte-Carlo draws required, and the computational cost of each draw. With fewer than 25 alternatives (JT), the results suggest that it is more advantageous to use the probit model (MNP) compared to the DKL. There appears to be little advantage in applying the kernel logit formulation relative to the MNP to cross-sectional data with a few alternatives. Regarding computational a ccuracy, the numerical results suggest that the parameter estimates of both models (MNP and DKL) are comparable and close to the true values from which the data sets were generated. However, both DKL and MNP formulations may lead to the maximization of a nonconcave objective function, resulting in flat log-likelihood functions, and identification problems. © 2005 INFORMS.

This paper investigates the allocation of household individuals to out-of-home maintenance activities using the rich activity-travel diary data from the San Francisco Bay Area. Two inter-related decisions are considered in this context: (i) whether the given activity episode is performed individually (solo) or jointly, and (ii) the person who participates in the activity, if it is a solo activity. To account for the conditional nature of the solo activity person selection, a nested mixed logit modeling framework is proposed and implemented to jointly analyze person allocation for all maintenance activities performed by a household on a given day. The model is used to investigate within-household effects and between-household differences. The proposed model relaxes some important restrictions in person allocation models by accounting for various sources of correlations and relaxing the assumption of constant variance across households. The proposed model is used to analyze the differences in person allocation between different types of households. The results indicate that life-cycle and household role, income, gender, employment status, and several types of constraints (activities including cost, time-availability, vehicle-availability, coordination constraints, and child-care obligations) affect person allocation decisions in the context of maintenance activities. The empirical results indicate the presence of various sources of correlations across persons, over activities, and within-household that are significant. In addition, the data also provides evidence that the unobserved variances in person selection utilities are not constant across households. A better understanding of these within-household interactions and between-household differences may be used in activity-based simulation models and to develop more effective and focused demand management measures. © Springer 2005.

Mobile phones are indispensable and ubiquitous tools that afford unprecedented levels of connectivity and accessibility to millions of users. A study investigated the influence of mobile phones on three travel-related dimensions: unplanned activity chaining, unplanned rideshares arranged by using mobile phones, and shopping by phone. These dimensions were investigated by using data from 400 workers in the city of Chennai, India. The results reveal that mobile phones significantly affect not only these travel dimensions but also activity participation. The data also provide evidence that social connectivity, activity characteristics, mobile phone use, and travel patterns are all strongly interlinked. Individual characteristics, such as flexible time and duration of working hours, and personal and household characteristics, such as age, gender, and vehicle availability, were found to be influential. The impact of mobile phones on the dimensions of unplanned stop making, ridesharing, and shopping trip substitution can have important practical implications for mode choice modeling, vehicle occupancy increase measures, and congestion alleviation measures.

This research investigated within-day dynamics and variations in mode choice within and across individuals at the activity-episode level. Specifically, dynamics and variability were examined at four levels: (a) effect of variation over time of explanatory attributes such as episode-trip characteristics, (b) influence of state-dependence (past choice decisions) on current choice, (c) role of habit persistence (i.e., the influence of lagged explanatory variables), and (d) the effect of unobserved preference heterogeneity that persists over repeated choices of individuals. A mixed-logit model was calibrated by using the rich activity-diary data from a 2000 activity-travel survey for the San Francisco Bay Area in California. Several alternative specifications of dynamics were tested. The aforementioned sources of variability and dynamics were found to be significant, and the empirical results showed a substantial improvement in model fit with the inclusion of these factors. The within-day dynamic model also outperformed alternative models in validation tests. The proposed model has important implications for improved activity-based demand models, travel demand forecasting, and evaluation of transportation control measures.