This paper deals with fuel consumption prediction based on vehicle velocity, acceleration, and road slope time series inputs. Several data-driven models are considered for this purpose, including linear regression neural network-based ones. The emphasis is accounting the impact when forming model inputs, in order to improve accuracy. A particular focus devoted conversion of length-varying driving cycles into fixed dimension inputs suitable networks. proposed algorithms parameterized tested...

A deep neural network-based approach of energy demand modeling electric vehicles (EV) is proposed in this paper. The model-based prediction based on driving cycle time series used as a model input, which properly preprocessed and transformed into 1D or 2D static maps to serve input the network. Several feedforward network architectures are considered for application along with different formats. Two models derived, where first one predicts battery state-of-charge fuel consumption at...

City bus transport electrification has a strong potential of improving city air quality, reducing noise pollution and increasing passenger satisfaction. Since the operation is rather deterministic intermittent, driving range- charging-related concerns may be effectively overcome by means fast charging at end stations and/or slow in depot. In order to support decision making processes, simulation tool for planning been developed it presented this paper. The designed use real/recorded cycles...

The authors of this paper propose a Markov-chain-based method for the synthesis naturalistic, high-sampling-rate driving cycles based on route segment statistics extracted from low-sampling-rate vehicle-tracking data. In considered case city bus transport system, segments correspond to sections between two consecutive stations. include lengths and maps average velocity, station stop time, station-stopping probability, all given along day an hourly basis. process cycle synthesis, transition...

The paper deals with a detailed analysis of game theory-based vehicle-pedestrian interaction model adopted from literature. For the sake simplicity only single-vehicle/single-pedestrian is considered. extended stochastic component that accounts for variability in pedestrian/driver behaviors and influence multi-agent environment. generalized scenarios where pedestrian appears at different distances relative to crosswalk. study organized as follows. First, an off-line model-derived probability...

This paper presents the synthesis and validation of multidimensional driving cycles represented by vehicle velocity, acceleration, road slope profiles. For this purpose, a rich set city bus has been recorded. First, Markov chain model is established based on time derivative states. Next, large synthetic generated using corresponding 8D transition probability matrix, which implemented in sparse form dictionary keys to improve computational efficiency reduce memory requirements. In support...

The paper firstly proposes a deterministic deep feedforward neural network model aimed at predicting the city bus velocity profile over receding time horizon based on following inputs: actual vehicle position, or short-term history of velocities, day and week. A systematic analysis influence different input subsets, interval length prediction is carried out to find an optimal configuration NN inputs hyperparameters. Secondly, stochastic version proposed, which predicts expectations standard...

Driving cycles reflect driver behavior and local traffic characteristics, they are often described in terms of vehicle velocity time profile widely used for different certification purposes. In order to obtain reliable fuel consumption estimates, realistic driving including the road slope should be used, especially electric vehicles whose energy may significantly vary depending on conditions. As a continuation previous research related statistical analysis rich set city bus cycles, this...

This paper proposes a neural network prediction model related to pedestrian crossing decisions, which is aimed be employed within autonomous vehicle safe speed control strategies. The of stochastic nature in order capture the inherent uncertainty and variability typically present real behavior. Instead predicting whole trajectory, only ego-vehicle control-related quantities are targeted for prediction, include entry exit time to/from area. To this end, two independent feedforward submodels...

This paper proposes a static, stochastic, deep feed-forward neural network-based model for prediction of city bus velocity along regular route. The emphasis is on proper formation outputs to consistently learn the conditional probability distribution vehicle based position as only input feature. First, rich set recorded driving cycles representative fleet ten buses statistically analyzed. Next, dataset properly downsampled and used computationally-efficient training validation network....

<div class="section abstract"><div class="htmlview paragraph">Driving cycles are usually defined by vehicle speed as a function of time and they typically used to estimate fuel consumption pollutant emissions. Currently, certification driving mainly for this purpose. Since artificially generated, the resulting estimates analyzes can generally be biased. In order address these shortcomings, recent research efforts have been directed towards development statistically representative...