Performance of current eco-routing methods

Abstract

Eco-routing is a vehicle navigation method that aims to minimize fuel or energy consumption for a given trip. It is based on a hypothesis that we can trade extra travel time for lower consumption. While the hypothesis was experimentally verified the design of a method that would fully exploit its potential proves challenging. Current solutions hinge on assumption that energy spent on any given road does not change in time. We challenge validity of this assumption by studying performance of such methods in detailed second-bysecond simulation that pronounces the time-dependencies. This allows us to quantify the real savings attainable with current eco-routing. I. INTRODUCTION Eco-routing 1 emerged as one of the strategies that aim to lower vehicle operating costs [12], [6], [4], [2]. The idea is to minimize energy (or fuel) consumption by route selection: given some origin and destination, eco-routing plots a route such that energy (fuel) needed to finish the trip is minimal. The routing is usually done on a graph where nodes represent junctions, edges represent roads and costs are estimated energies needed to travel between two junctions the road connects. Minimal path routing 2 can then be used to find the route that minimizes total energy for the trip. Authors typically reduce the complex time-variant functions that describe the costs. They must be time-invariant and nonnegative in order to use Dijkstra's routing algorithm, which is a common choice between authors. Validation is often done using the same assumptions. Full experimental validation would require a host of identical vehicles to depart from spatially and temporally identical place in order to measure consumptions on different paths to destination reliably.

Ecological traffic management: A review of the modeling and control strategies for improving environmental sustainability of road transportation

“…Probabilistic approaches are a solution in case of lack of information about the vehicle dynamics. By construction, they are less accurate than models based on the actual speed, but can be effectively used to estimate emissions and energy consumption (Kubička et al, 2016). To improve these methods, traffic prediction models could be integrated to determinev (cf.…”

Efficient Mobility-on-Demand System with Ride-Sharing

“…By considering link-based variables, they can address driving heterogeneity, thus are more accurate than macroscopic models. However, most of the existing mesoscopic models for eco-routing use parametric regression-based models [12,13] or power balance models [14][15][16] and are not accurate enough due to the complexity of traffic scenario and nonlinearity of vehicle powertrains. Advanced data-driven methods such as support vector machines (SVM) [17], neural networks (NN) [18] and multivariate adaptive regression spline (MARS) [19] were also studied, and many outperformed the traditional methods.…”

Eco-Routing of Plug-In Hybrid Electric Vehicles in Transportation Networks

“…Several routing algorithms have been proposed in the literature for conventional vehicles which are capable of finding the energy-optimal paths using historical and online traffic data [4]- [8]. Kubicka et al [9] performed a study to compare the objective values proposed in the eco-routing literature and showed that the performance of eco-routing algorithms is highly dependent on the method used to calculate the traveling cost of each link. Although eco-routing of conventional vehicles is well studied, there is little research that addresses the case of PHEVs [10].…”