Towards drivers’ safety with multi-criteria car navigation systems

https://doi.org/10.1016/j.future.2022.04.019Get rights and content

Highlights

  • Analysis of the impacts of a multi-criteria route selection in route features.

  • Four different techniques to measure route features.

  • Two novel route selection methods.

  • A novel dataset with routes information and GPS coordinates.

Abstract

Usual car navigation systems are configured to propose either the shortest or the fastest path between any origin–destination pair, neglecting the particularities of the territory. Some roads are impracticable when raining, some others are to avoid at night for the scarce lighting, or less safe for the presence of criminality and high accident ratio. On the other hand, longer paths can be safer and more pleasant as they pass through less noisy zones, with the presence of beautiful landscapes. In this paper we analyze the faults in current car navigation systems, especially quantifying the trade-off between safety and traveling time or path length. We propose two multi-criteria route planning methods, HVT (Hierarchical with Variable Tolerance) and R2V (Route to Vector), suggesting the best path to drivers also considering safety or multiple drivers’ specific needs. A dataset of 3,170 paths from 600 origin/destination pairs within London is created and shared to the research community. With this dataset, we show that selecting routes with reduced driving risks is indeed possible with a marginal increase in travel times.

Introduction

The navigation system market, including automotive applications, is expected to register a CAGR of 11.3% over the forecast period 2020–2025 [1]. In particular, car navigation systems contain digital maps, with information about the neighboring areas, and use route planning algorithms to give directions to drivers. These route planning algorithms tend to focus on finding the fastest or the shortest path between origin–destination (OD) pairs [2].

The downside of the single-criterion approach, however, is the chance to lead drivers through unpleasant or even dangerous routes without significant time-related gains. Other characteristics may have an impact on drivers’ overall satisfaction when traveling as well. The contact with nature has been found to have a positive effect on mental health and wellness, even if just visual, as is the case for drivers [3]. On the other hand, traffic may increase stress levels and even mortality due to noise and air pollution [4]. Besides health-related problems, drivers also experience anger issues on road, that influences the drivers’ attitude, with an increase in aggressive driving, risky driving, driving errors, and accidents [5]. Choosing alternative routes, therefore, can have an enormous impact on the drivers’ safety [6], [7], [8], [9].

Fig. 1 shows an example of a set of routes for a given OD within the London area. The map includes an overlay heatmap representing the criminality rates and a second overlay indicating locations of vehicle accidents with relative severity. Route 1, represented as a bold yellow path, is the fastest route and it is recommended by most car navigation systems. It passes through an area with high criminality rate though. On the other hand, Route 2, represented by a red line, goes through an area with less criminality and more nature, and could be preferred over the first route, especially if the duration of both routes is similar.

This paper aims to provide methods to measure different route characteristics that may influence the driver’s safety and pleasantness perception. We evaluate different route selection methods with either a single or a multiple criteria approach, with the intent to find the best path from a set of previously calculated routes. We introduce two novel multi-criteria route selection methods. The first method, called Route to Vector (R2V), translates route features into vectors and finds the one closest to the best vector. The second, called Hierarchical with Variable Tolerance (HVT), follows a user-defined feature order to reach the best route.

To evaluate our approaches, we have beforehand built a novel dataset containing criminality, traffic, accidents, nature, tourist attractions, and trajectory information data about 3170 routes from the city of London [10]. Safety-related parameters are individually evaluated with the proposed methods to emphasize safety gains. The obtained results show that our multi-criteria approaches decrease driving risks without significant time-related penalties. Also, we confirm that relying just on duration or path length may lead to dangerous and unpleasant trajectories. Throughout this paper, the names trajectory, route and path are used interchangeably.

The remainder of this paper is structured as follows. Section 2 provides an overview of car navigation systems, while in Section 3 we examine previous works on route classifications. Next, in Section 4, we describe our collected dataset containing a set of route parameters that can influence the driver’s satisfaction. Then, we propose two route selection methods, the Route to Vector and the Hierarchical with Variable Tolerance, based on multiple criteria in Section 5. We evaluate the performance of both proposed methods to assess route classification and drivers’ security improvements in Section 6. Finally, Section 7 concludes this paper and draw future directions.

Section snippets

Car navigation systems overview

Car navigation systems are in-car or smartphone based systems implemented to aid drivers in planning a route for a given OD pair. Upon receiving the OD pair translated into coordinates of a digital map, these systems compute and advise a set of possible routes, usually considering duration as the preferential single criterion. Car navigation systems can also have additional features such as real-time alerts about traffic events. However, this functionality can be viewed as an add-on available

Related work on multi-criteria planning

In this section, we present an overview of recent navigation systems that provide different route planning methods, taking into account the notion of multi-criteria car navigation systems. These systems are complex and do not exclude real-time, predictive, and multi-modal planning. The multiple criteria approach may be implemented in different ways. We present different methods with three different objectives: the first and second ones focus on tourist needs, the third one is directed to

Dataset for multi-criteria planning

To compare our two route selection proposals with the baseline single-criterion approach, we built a dataset composed by several alternative paths for the same OD pairs within the city of London [10]. The dataset contains 3170 paths from 600 random OD pairs. Each OD pair creates a route set that contains a varying number of partially or totally disjoint routes generated using the HERE Maps API [18], along with duration estimations. The HERE Maps API does not return paths exceeding the fastest

Multi-criteria classification methods

The proposed multi-criteria methods follow a workflow composed of: (i) per route feature calculation, (ii) feature normalization, and (iii) routes ranking. The input is a set of routes R that can be randomly selected or previously calculated considering any preliminary metric of interest. Hence, each route Ri is described by a n-tuple f considering features of interest. We can select any possible subset computed using any combination of features available. In this paper, we use path duration to

Multi-criteria models evaluation

In this section, we analyze the feature space beforehand, then we evaluate the proposed methods for multi-criteria route selection with respect to single criterion approaches. Finally, we highlight security gains and aspects that could be emphasized using our multi-criteria contributions.

The values presented in this section were gathered from three different sources. HERE Maps API provided route data from the random OD coordinate pairs, such as path coordinates, trip duration with and without

Conclusion

In this paper, we focus on route selection improvements considering alternative paths beyond the fastest one. In particular, we stress that the fastest (or even the shortest) path proposed by car navigation systems may present a higher level of safety risks, while slightly slower (or longer), well-selected paths, could reduce such risks. Thus, we propose and analyze two multi-criteria route selection methods, HVT (Hierarchical with Variable Tolerance) and R2V (Route to Vector), and four

CRediT authorship contribution statement

Leonardo Solé: Software, Investigation, Data curation, Writing - original draft, Visualization. Matteo Sammarco: Conceptualization, Validation, Writing - review & editing, Project administration. Marcin Detyniecki: Resources, Funding acquisition. Miguel Elias M. Campista: Methodology, Formal analysis, Writing - review & editing, Supervision, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This paper was partially supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001. It was also supported by CNPq, Brazil, FAPERJ, Brazil Grants E-26/211.144/2019 and E-26/202.689/2018, and FAPESP, Brazil Grant 15/24494-8.

Leonardo Solé Rodrigues received his Social Communications degree from the Superior School of Marketing and Propaganda (ESPM), Brazil, in 2011, and is currently pursuing his Electronic and Computer Engineering degree from the Federal University of Rio de Janeiro (UFRJ). He is part of the Group of Teleinformatics and Automation Group (GTA) from UFRJ, working with computer networking and machine learning.

References (22)

  • CarlosC.H. et al.

    Towards participatory sensing of regions of interest with adaptive sampling rate

    Veh. Commun.

    (2020)
  • Leonardo Solé Rodrigues received his Social Communications degree from the Superior School of Marketing and Propaganda (ESPM), Brazil, in 2011, and is currently pursuing his Electronic and Computer Engineering degree from the Federal University of Rio de Janeiro (UFRJ). He is part of the Group of Teleinformatics and Automation Group (GTA) from UFRJ, working with computer networking and machine learning.

    Matteo Sammarco received his the B.Sc. and M.Sc. degrees in Computer Engineering from Università degli Studi di Napoli Federico II, Italy, in 2008 and 2010, respectively. In 2014 he received the Ph.D. in Computer Science from Sorbonne Université, France. He has worked as researcher at Télécom ParisTech and Laboratory of Information, Networking and Communication Sciences, both in France. Currently, he is research data scientist at Axa Group Operations, mainly working on machine learning applied to Safety, Smart Mobility and IoT.

    Marcin Detyniecki studied mathematics, physics and computer science at the Sorbonne Université Paris. In 2000 he obtained his Ph.D. in Artificial Intelligence from the same university. Between 2001 and 2014, he was a research scientist of the French National Center for Scientific Research (CNRS). He has been researcher at the University of California at Berkeley and at Carnegie Mellon University (CMU) and visiting researcher at the University of Florence and at British Telecom Research labs.

    He is currently Group Chief Data Scientist and Global Head of Research at AXA Group Operations. His work focuses on Machine Leaning, Artificial Intelligence, Computational Intelligence, Multimedia Retrieval, Fair and Transparent AI.

    Miguel Elias M. Campista received a D.Sc. in Electrical Engineering from the Federal University of Rio de Janeiro (UFRJ), in 2008. He is an associate professor in the Electronic and Computer Engineering Department at Poli/UFRJ and a full professor in the Electrical Engineering Program (PEE) at COPPE/UFRJ. Prof. Campista is an Affiliate Member of the Brazilian Academy of Science (ABC), is on the board of directors of the Brazilian Computer Networking Lab, is an Editor of the Annals of Telecommunications Journal, and an IEEE senior member. His research interests are on data and network science.

    View full text