Detecting the adherence of driving rules in an energy-efficient, safe and adaptive driving system

Highlights

• A rule matching algorithm that matches driving rules against a driving behavior is proposed.

• Development of the rule matching algorithm in an energy-efficient, safe and adaptive driving system.

• Evaluation of the rule matching algorithm using 15 journeys and three driving rules.

• Results show a higher performance than existing algorithms.

Abstract

An adaptive and rule-based driving system is being developed that tries to improve the driving behavior in terms of the energy-efficiency and safety by giving recommendations. Therefore, the driving system has to monitor the adherence of driving rules by matching the rules to the driving behavior. However, existing rule matching algorithms are not sufficient, as the data within a driving system is changing frequently. In this paper a rule matching algorithm is introduced that is able to handle frequently changing data within the context of the driving system. 15 journeys were used to evaluate the performance of the rule matching algorithms. The results showed that the introduced algorithm outperforms existing algorithms in the context of the driving system. Thus, the introduced algorithm is suited for matching frequently changing data against rules with a higher performance, why it will be used in the driving system for the detection of broken energy-efficiency or safety-relevant driving rules.

Introduction

When saving energy and protecting the environment became fundamental for politics and society, several laws were enacted to reduce the greenhouse gas emissions, like the CO₂ limitations, for passenger cars in the European Union. Another factor that became important during the past few decades is road safety, as the increasing number of cars led to more accidents and fatalities on the road. On the basis of the enacted laws with the goal to save the environment (Wessellink, Harmsen, & Eichhammer, 2010) and the increasing importance of road safety, car manufacturers are trying to optimize the car and its individual parts like the car body, the engine, or the power train. Furthermore, new methods are being invented to increase the energy-efficiency and safety of cars, like the regenerative brake (Patil, 2012), which converts the kinetic energy during a brake application to electric energy or the anti-lock brake system (Burton, Delaney, Newstead, Logan, & Fildes, 2004), which improves safety by preventing the wheels from locking up.

Besides the optimization of the car itself, there is also the potential to increase the energy-efficiency and improve road safety by adapting the individual driving behavior to the given driving situation. The studies (Xiaoqiu, Jinzhang, & Guoqiang, 2011) and (Chin & Quek, 1997) revealed that the driving behavior has an effect on road safety. This has also been verified in the accidents report of the German Statistical Office (German Statistical Office, 2014), which showed that 86% of the accidents with damage to persons in Germany in 2013 happened because of driver mistakes, see Fig. 1. Several studies have shown that energy savings up to 30% are possible with the adaptation of the driving behavior (Haworth, Symmons, 2001, Helms, Lambrecht, Hanusch, 2010, van Mierlo, Maggetto, van de Burgwal, Gense, 2004). However, according to Bongard (2007) this 30% savings is only possible with experienced drivers. The energy savings vary depending on the given driving practices. In short-term driving practices, like energy-efficiency training or contests, an energy saving of about 24% is possible as shown in the tests conducted by the car manufacturer Ford (Spencer, 2008). In contrast, Barkenbus (2009) calculates the energy savings in sustained driving practices at 5% when the drivers have no continuous energy-efficiency related feedback after the initial training. Continuous feedback is defined by showing the driver energy-efficient relevant driving recommendations every day. However, with continuous feedback, the energy savings is 10%.

There are already driving systems and methods which try to reduce energy consumption or to improve road safety at the vehicle level, such as the driving systems of Khayyam, Nahavandi, and Davis (2012) and Milanes, Perez, Godoy, and Onieva (2012) and the method of Jo, Lee, Park, Kim, and Kim (2014). Furthermore, there are also driving systems, such as those of Fiat (2010), Cho (2008) or Lotan and Toledo (2006), which focus on improving the driving behavior in terms of energy-efficiency or safety. However, these driving systems only cover either the area of energy-efficiency or safety and provide insufficient feedback to the driver, for example by showing a green lamp when the driver is driving in an energy-efficient manner. Furthermore, they do not adapt to the individual driver, by considering the driver condition or the individual driving behavior. In contrast, the adaptive and rule-based driving system that is in development has the goal to improve the driving behavior in terms of energy-efficiency and safety by giving customized recommendations to the driver during a journey. Thus, according to Barkenbus (2009), this can lead to an increase in the energy-efficiency of up to 10%. The driving system adapts to the driver by customizing the recommendations on the basis of his or her reaction to the previously given recommendations and the driver condition. For example, the driving system creates no recommendation when the driver is under stress. Furthermore, it decreases the generation frequency of a recommendation when the driver has ignored a recommendation several times. The energy-efficiency and safe driving behavior is described by a set of rules. These rules are used to check if the driver is driving efficiently and safely. If the breaking of a driving rule is detected, an energy-efficient or safety-relevant recommendation is shown to the driver.

Rule matching algorithms, such as Rete (Forgy, 1982) or Treat (Miranker, 1987), are designed to check a certain data set against rules, making them suitable for checking driving rules against driving behavior in order to find an energy-inefficient or unsafe driving behavior. However, existing rule matching algorithms, like Rete or Treat, are not designed to handle frequently changing data like in a driving system, which is why the performance of these algorithms is not ideal for their use in driving systems. There are already improved versions of the Rete algorithm, like the algorithm of Li, Liu, Cao, Yin, and Yao (2016). They created an algorithm based on Rete that allows to subtask the task of rule matching to different computers in a distributed or parallel computing environment. Therefore, the algorithm decomposes rules to sub rules and distributes them to a distributed environment for parallel match, after which the algorithm merges the results through a reduce function. According to Li et al., this approach improves the match efficiency for massive rules matching. Another modified Rete algorithm is introduced in Liu, Huang, Zhang, and Du (2014), who try to improve the efficiency of the Rete algorithm for the area of social insurance audit using the active hash method. The active hash method allows to improve the search time for a node within the network by managing the nodes and the corresponding facts within a hash table. The experimental results showed that the active hash method improves the efficiency of the Rete algorithm. Rete+ is another improved Rete algorithm introduced by Gao, Qiu, and He (2013) that has the focus on matching rules in the context of Web of Things environments, like smart homes. According to Gao et al. the Rete algorithm executes the same rules in the environment of Web of Things, why the Rete+ algorithm stores the index of the alpha nodes for the last executed rules. Each fact that is put into the working memory, it will be matched first with the alpha nodes whose index were stored. In case of an successful match, the beta nodes of the corresponding alpha nodes are updated. However, when the matching fails, the algorithm goes on to match the facts like in the original Rete algorithm. The experimental evaluation showed an higher performance of the Rete+ algorithm in comparison to the Rete algorithm in the context of smart home environment.

However, the presented improved Rete algorithms are designed for specific environments, like Web of Things, social insurance and so on. They are using parallel computing, active hash methods or storing the last used alpha nodes to optimize the Rete algorithm. However, these optimizations are not sufficient for the environment of the driving system, as the optimizations are fitted for a specific environment that is different from the environment of the driving system. Furthermore, the driving system has to match driving rules against data that is updated in near time. The goal of the rule matching algorithm introduced in this paper is to match driving rules to a data set with a higher performance than existing rule matching algorithm in the context of the driving system, in order to allow the driving system to show recommendations to the driver in time. An early stage of the introduced rule matching algorithm is explained in Yay, Martńez Madrid, and Ortega Ramírez (2014).

This article presents an adaptive driving system for energy-efficient and safe driving and focuses on the development of a rule matching algorithm that checks driving rules against the driving behavior to find an energy-inefficient or unsafe driving behavior. As the rule matching algorithm is in the context of a driving system, its architecture is explained in Section 2. Available rule matching algorithms, like Rete, Treat or Leaps, are described in Section 3. Section 4 introduces the concepts and the algorithm of the developed rule matching algorithm. The evaluation of the developed rule matching algorithm, on the basis of 15 journeys, is described in Section 5. The corresponding results of the evaluation are presented in Section 6. Finally, the conclusion and further work based on this proposal will be discussed in Section 7.