Stephen D. Boyles

Auction-based autonomous intersection management

Autonomous vehicles present new opportunities for addressing traffic congestion through flexible traffic control schemes. This paper explores the possibility that auctions could be run at each intersection to determine the order in which drivers perform conflicting movements. While such a scheme would be infeasible for human drivers, autonomous vehicles are capable of quickly and seamlessly bidding on behalf of human passengers. Specifically, this paper investigates applying autonomous vehicle auctions at traditional intersections using stop signs and traffic signals, as well as to autonomous reservation protocols. This paper also addresses the issue of fairness by having a benevolent system agent bid to maintain a reasonable travel time for drivers with low budgets. An implementation of the mechanism in a microscopic simulator is presented, and experiments on city-scale maps are performed.

Intersection Auctions and Reservation-Based Control in Dynamic Traffic Assignment

Autonomous vehicle (AV) technology is maturing, and AVs are being test-driven on public roads. A promising intersection control policy, tile-based reservation (TBR), proposed by Dresner and Stone in 2004, could improve intersection capacity beyond the capabilities of optimized traffic signals. Although TBR has been studied in several microsimulation models, it has yet to be analyzed under user equilibrium behavior. In this study, TBR was modeled in the dynamic traffic assignment to draw on the extensive literature on vehicle routing behaviors. With the proposed model, TBR can be computationally simulated on large city networks, with the goal of solving the traffic assignment problem. TBR also arbitrarily prioritizes vehicle movement, and high-value-of-time travelers may be able to gain priority through intersection auctions, as suggested by the literature. An in-depth study of simple intersection auctions found that much of the benefit (over first-come, first-served prioritization) resulted from the randomizing effect of auctions giving larger queues of vehicles greater shares of the intersection capacity.

A general framework for modeling shared autonomous vehicles with dynamic network-loading and dynamic ride-sharing application

•We present a framework for modeling shared autonomous vehicles (SAVs) compatible with a general class of traffic simulation.•We implement our framework using a cell transmission model simulator on a city network.•SAVs can greatly increase congestion, and therefore road congestion should be included in all SAV models.•We compare polynomial-time heuristics for dynamic ride-sharing and preemptive relocation for SAVs.

Robust Optimization Strategy for the Shortest Path Problem under Uncertain Link Travel Cost Distribution

This article showed how numerical experiments conducted on small to large networks compare the robust optimization-based strategy to the classical deterministic shortest path in terms of the uncertainty. A robust optimization approach for the shortest path problem where travel cost is uncertain and exact information on the distribution function is unavailable is developed. The article showed that under such conditions the robust shortest path problem can be formulated as a binary nonlinear integer program, which can then be reformulated as a mixed integer conic quadratic program. This article presented an outer approximation algorithm as a solution algorithm, which is shown to be highly efficient for this class of programs.

Network Model for Rural Roadway Tolling with Pavement Deterioration and Repair

A rural pricing model which calculates diversion endogenously with a network assignment model is described in the article. Roadway tolling is often tied to revenue generation for roadway maintenance in rural areas and rural pricing models should directly incorporate a pavement deterioration and maintenance model. However, the interactions between these models are not simple, because tolls cause traffic diversion, which in turn affects deterioration rates and forecasted revenue. The pricing model presented in this article captures deterioration rates and pavement condition in the toll-setters objective function, which maximizes long-run net present value of the highway infrastructure. A demonstration is made using a network that represents the state of Wyoming (28 zones, 60 nodes, and 188 links). A novel deterioration model is used which is particularly suitable for computational efficiency and the resulting model is discontinuous and nondifferentiable and involves solving a knapsack problem as a subproblem. Therefore, a simulated annealing-based algorithm is presented to solve it, in the framework of a new solution method built upon partitioning the feasible region. Sensitivity analyses reveal that although the locations for optimal tolling are relatively stable as demand changes, the revenue collected can substantially vary. Future research should investigate strategies for incorporating more advanced pavement network models. This paper used simple models for computational reasons.

Demand Profiling for Dynamic Traffic Assignment by Integrating Departure Time Choice and Trip Distribution

One challenge in dynamic traffic assignment (DTA) modeling is estimating the finely disaggregated trip matrix required by such models. In previous work, an exogenous time distribution profile for trip departure rates is applied uniformly across all origin-destination (O-D) pairs. This article develops an endogenous departure time choice model based on an arrival time penalty function incorporated into trip distribution, which results in distinct demand profiles by O-D pair. This yields a simultaneous departure time and trip choice making use of the time-varying travel times in DTA. The required input is arrival time preferences, which can be disaggregated by O-D pair and may be easier to collect through surveys than the demand profile. This model is integrated into the four-step planning process with feedback, creating an extension of previous frameworks which aggregate over time. Empirical results from a network representing Austin, Texas indicate variation in departure time choice appropriate to the arrival time penalties and travel times. Our model also appears to converge faster to a dynamic trip table prediction than a time-aggregated coupling of DTA and planning, potentially reducing the substantial computation time of combined planning models that solve DTA as a subproblem of a feedback process.

Network Route Choice Model for Battery Electric Vehicle Drivers with Different Risk Attitudes

Research on the range anxiety of battery electric vehicle (BEV) drivers is limited, and research on the route choice of such drivers has been restricted to a fixed range limit modeled as a distance constrained shortest path problem. In this paper, a more general network route choice model based on the range anxiety of BEV drivers is formulated as a nonadditive shortest path problem. More appropriate for BEVs, a tour-based analysis with a continuum of range limits is considered, and an outer approximation algorithm has been used to find the optimal path. Numerical experiments on a small network demonstrate how the routes taken by BEV drivers are influenced by their risk attitudes and uncertainty in the predicted range of the vehicle.

Modeling Parking Search on a Network by Using Stochastic Shortest Paths with History Dependence

A substantial amount of urban traffic is related to drivers searching for parking. This study developed an online stochastic shortest path model to represent the parking search process in which drivers must choose whether to park at an available space or continue searching for a space closer to their destination. Existing online shortest path algorithms had been formulated for the full-reset or no-reset assumptions on revisiting links. As described in this paper, neither assumption was fully suitable for the parking search process. Accordingly, this paper proposes an asymptotic reset model that generalizes the full-reset and no-reset cases and uses the concept of reset rate to characterize the temporal dependence of parking probabilities on earlier observations. In this model, drivers try to minimize their expected travel cost, which includes the driving cost and the cost of walking from a parking spot to the actual destination conditioned on the parking availability on m most recently traversed links. The problem was formulated as a Markov decision process and was demonstrated with a network representing the neighborhood of the University of Wyoming campus in Laramie. The case study successfully shows the extra time used by drivers to cruise for an acceptable parking space and illustrates the impact of m on the computation effort required to compute an optimal policy.

Increasing Evacuation Communication Through ICTs: An Agent-based Model Demonstrating Evacuation Practices and the Resulting Traffic Congestion in the Rush to the Road

Abstract Understanding evacuation practices and outcomes helps crisis and disaster personnel plan, manage, and rebuild during disasters. Yet the recent expansion in the number of information and communication technologies (ICTs) available to individuals and organizations has changed the speed and reach of evacuation-related messages. This study explores ICTs’ influences on evacuation decision-making and traffic congestion. Drawing from both social science and transportation science, we develop a model representative of individual decision making outcomes based on the amount of ICT use, evacuation sources, and the degree of evacuation urgency. We compare the evacuation responses when individuals receive both advance notice of evacuation (ANE) and urgent evacuation (UE) messages under conditions of no ICTs and prolific ICT use. Our findings from the scenarios when there is widespread ICT use reveal a shift in the evacuation time-scale, resulting in traffic congestion early in the evacuation cycle. The effects of this congestion in urgent situations are significantly worse than traffic congestion in the advance notice condition. Even under conditions where face-to-face communication is the only option, evacuations still occur, but at a slower rate, and there are virtually no traffic congestion issues. Our discussion elaborates on the theoretical contributions and focuses on how ICTs have changed evacuation behavior. Future research is needed to explore how to compensate for the rush to the road.

Quantifying distributions of freeway operational metrics

Abstract Nonrecurring congestion creates significant delay on freeways in urban areas, lending importance to the study of facility reliability. In locations where traffic detectors record and archive data, approximate probability distributions for travel speed or other quantities of interest can be determined from historical data; however, the coverage of detectors is not always complete, and many regions have not deployed such infrastructure. This paper describes procedures for estimating such distributions in the absence of this data, considering both supply-side factors (reductions in capacity due to events such as incidents or poor weather) and demand-side factors (such as daily variation in travel activity). Two demonstrations are provided: using data from the Dallas metropolitan areas, probability distributions fitting observed speed data are identified, and regression models developed for estimating their parameters. The application of the demand-side procedure is seen to improve the accuracy of the prediction. The second demonstration, using data from the Seattle metropolitan area, identifies the appropriate capacity reduction applied to planning delay functions in the case of an incident.