A systematic literature review of vehicle speed assistance in intelligent transportation system

Due to the emergence of new technologies over the past decade, vehicle speed assistance systems in intelligent transportation systems have frequently been discussed. Up to now, a systematic literature review has not been presented to discover and evaluate the different vehicle speed assistance approaches for on-road vehicles in intelligent transportation systems. To overcome this issue, this research identiﬁed peer-reviewed articles published in the most well-known libraries from 2011 to 2020. 79 primary studies were then projected and a systematic analysis of the selected literature was conducted. The ﬁndings show different driving goals, namely eco-driving, safety, comfort, travel time improvement as well as the high-level objectives addressed by vehicle speed assistance systems. The analytical discussions are provided to show different perspectives, properties and limitations of the existing solutions. This analysis allows to provide future challenges and directions in this ﬁeld of research.


INTRODUCTION
Vehicles have always been a part of human life, and the population has a direct effect on the increase in vehicles, which ultimately affects the environment in terms of pollution and safety. A recent investigation of the total number of on-road vehicles conducted by [1] indicated that there were over 1 billion vehicles all over the world in 2010. It has been estimated that over 2 billion passenger cars travel the streets and roads of the world today, and the number of vehicles is expected to double by 2050 [2], demanding the capacity much beyond the present level of roadways. As the demand increases beyond the capacity, there is a necessity to overcome the negative impacts of the increasing number of on-road vehicles, namely congestion, pollution, and accidents, ultimately affecting society's social, economic, and environmental aspects of lives. In this regard, intelligent speed assistance systems have shown many advantages in decreasing the aforementioned adverse transportation effects. Most notably, the objectives of vehicle speed assistance encompass a range of traffic scenarios such as highways and intersections and can serve various purposes depending on the This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. IET Intelligent Transport Systems published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology driving goals. Such vehicle speed assistant systems have been considered as having high potential in providing road safety to vehicle passengers [3]. Despite these systems' assistive nature, they can also work in mandatory mode to regulate the traffic speed limit and collaborate with the law enforcement section of Intelligent Transportation Systems (ITS). Point-to-point (P2P) speed enforcement is a high-tech method that can work with current mandatory speed assistance systems for road regulation and driver satisfaction. P2P involves calculating the average speed along a road section using a series of cameras at multiple locations to decrease the unsafe states caused by sudden brakes, which might occur in traditional single-point camera systems [4][5][6]. Hence, it is crucial to review the existing literature, especially related to vehicle speed assistance methods and their applicability. Therefore, to identify the solutions that have been provided for speed assistance, it is essential to map out relevant papers and scholarly works systematically. This paper seeks to focus on existing literature concerning the different types of vehicle speed assistance systems, proposed solutions, and objectives, including green light speed advisory (GLOSA), intersection conflict avoidance, and traffic harmonisation as well as environmental/social goals. The objective is to provide a community-driven initiation for a better study of developing vehicle speed assistance techniques that explores the appropriate solutions concerning different traffic scenario limitations. Toward this goal, we will critically examine current works and studies on vehicle speed assistance and use our insights to develop new directions.

Prior research
To the best of our knowledge, there appears to exist a limited number of surveys and systematic literature reviews (SLRs) concerning the different types of vehicle speed assistance techniques. One of the recent survey papers in the realm of speed harmonisation was conducted by [7]. In this study, the authors have focused on presenting a review of traditional speed harmonisation methods based on variable speed limit (VSL) and ramp metering (RM) techniques. The authors examined the literature considering speed harmonisation algorithms, mechanisms and the effects of emerging technologies. They pointed out the scarcity of speed harmonisation for dynamics traffic flow and lack of applying emerging technology in their study. From our perspective, this survey gives valuable understandings to researchers who might be interested in the design and development of intelligent speed assistance systems using intelligent infrastructures. In contrast to this study, we eliminated traditional VSL studies and focused on speed assistance techniques and different speed advisory objectives in ITS. In a recently published paper in 2020, Mintsis et al. conducted a systematic review concerning Eco-driving approaches near intersections [8]. However, the design of this study seems to suffer a number of flaws. First, the authors have not provided any information regarding criteria and quality assessment for the selected studies, and the study lacks technical research questions which is the basis of any SLR. Second, they have considered only the Eco-driving aspect of vehicle speed assistance approaches and have not analysed ITS speed assistance approaches for different driving objectives or traffic scenarios specifically.
The studies discussed above have investigated only a subset of vehicle speed assistance techniques focusing a specific traffic scenario (intersection) with limited driving objectives in mind. The field of advising or recommending appropriate speed to connected vehicles (CV) still needs further investigation and insights. Therefore, it is necessary to provide a specific summary of the recent research papers, particularly in the realm of vehicle speed assistance techniques, to provide insight for future research directions.

Research goals
The research purpose is to analyse existing studies and their findings to give an outline of the efforts of vehicle speed assistance approaches in ITS. To focus on the research, we developed several technical research questions, as illustrated in Table 1. We will answer the research questions in Section 4.

Contribution and layout
This SLR is supplementary to existing studies and provides the following contributions for those who are interested in ITS and especially vehicle speed assistance in the fields of transportation and computer science to further their work: • We identified 79 primary studies considering vehicle speed assistance in ITS up to early September 2020. Other researchers can use this list of studies to base their work on in this specific field. • Next, we chose 50 primary studies that satisfy the criteria we have set for quality assessment phase. These studies can present rich data for comparative examination against similar research. • We then reviewed the data within 50 studies comprehensively and extracted the data to reveal ideas and concerns related to vehicle speed assistance systems. • We present a meta-analysis of vehicle speed assistance's methods and objectives to improve intelligent transportation systems and emerging technologies in this context. • We express the limitations and produce guidelines to support further investigation in this area.
The rest of the paper is structured based on [9], and the sections' order is as follows: Section 2 explains the methods with which the primary studies were systematically selected for analysis. Section 3 presents the findings of all the primary studies selected. Section 4 discusses the findings related to the research questions presented earlier and offers some suggestions for future research. We conclude our findings in Section 5.

RESEARCH METHODOLOGY
To perform an in-depth investigation and to answer the research questions, we conducted the SLR under the guidance published in [10]. We aimed to move through the planning, conducting and reporting phases of the review in iterations to allow for a thorough evaluation of the SLR.

Selection of primary studies
In this study, we highlighted the primary studies by a simple query to the particular publications or search engines. The search was conducted in September 2020, and can therefore be assumed to contain the most recent papers published in 2020 or earlier that would assist in answering the RQs. To ensure the relevance of the selected articles, the query string was restricted to run against paper titles, abstracts and keywords as follows: ("speed management" OR "speed advisory" OR "speed recommendation" OR "speed assistance" OR "speed assistant" OR "speed harmonisation" OR "speed guidance") The electronic databases in which searches were performed are listed in Table 2.
The initial Digital Libraries' searches resulted in 355 papers. Afterward, we filtered these search considering inclusion/exclusion criteria, which are illustrated in Section 2.2, and the analysis of title, abstracts, and keywords that resulted in the remaining 100 papers. We further analysed the remaining papers' full text against inclusion/exclusion criteria, resulting in 71 papers. Then, the produced set of results were candidates to run forward and backward snowballing as described in [11] in which references to and from those papers were analysed to check for additional literature. First, backwards snowballing was performed by going through the reference lists in all papers, Then, a forward snowballing was carried out, where we have analysed what other papers refer to members of the identified set. Figure 1 depicts the iterative process of primary studies filtering.

Inclusion and exclusion criteria
The studies that are to be included in this SLR must present a computerised approach in order to advise the speed of passenger cars and provide analytical results considering applications and objectives. They must have been published in peer-reviewed journals as described in 2.1 and written in English language. Additionally, the papers that focused on law enforcement and road restriction based speed management, such as bumps, were omitted. The key inclusion/exclusion criteria are described in Table 3.

Inclusion criteria Exclusion criteria
The paper provides analytical data related to the application and research objectives.
Papers that only evaluate and compare existing methods' performances.
Papers that are published in peer-reviewed journals.
Papers that only address the speed management issue concerning road physical restrictions and variable speed limits (VSL).
Paper considering on-road vehicles Literature such as technical reports or government documents.
Non-english papers.

Selection results
We found a total of 355 studies by the initial keyword searches against the title, abstract, and keywords on the well-known libraries, as mentioned in Table 2. After applying the inclusion/exclusion criteria on the initial papers set, the number of papers was reduced to 100. Furthermore, considering the criteria, we have read the remaining papers entirely, and after reapplying the criteria, 71 primary studies remained. Finally, the relevant papers underwent a forward and backward snowballing process, as indicated in [11], in which additional primary studies were targeted for inclusion in this SLR as 79.

Quality assessment
In order to assess the relevant papers to answer the research questions, The quality assessment of primary studies was done by the guidance set provided by [10]. Based on suggestion offered by Hosseini et al., [12] we chose five randomly papers to check their effectiveness throughout the quality assessment process: 1. Intelligent transportation system. The paper must provide speed assistance system considering ITS platform. 2. Context. To accurately judge the research, the paper must present enough context for the research objectives and experimental findings to answer research question RQ1. 3. Application. The study must provide enough information for the applicability of the proposed method to a specific traffic scenario, which will help in answering research question RQ2. 4. Road users context. The paper must consider passenger cars and not focus on public vehicles. 5. Solution and performance evaluation. The paper should give details of the evaluation environment and solution models, which will help answering questions RQ3 and RQ4.
Afterward, the provided checklist for quality assessment was applied to all other primary studies extracted during the search process. Figure 2 depicts the number of primary studies published by different publications included in our SLR during 2011-September 2020.

Data extraction
To assess data integrity to test the information record-keeping of selected studies, we have extracted data of every research article qualified during the assessment process. Before expanding the data extraction process to all selected studies, we examined the process for ten initial primary studies. As illustrated in Figure 1, we have shown the selection process of primary papers from a set of articles which were found during initial searches.

Data analysis
We organised the data from reviewed papers within the qualitative and quantitative categories to present the answers to research questions. Furthermore, we then conducted a systematic review of screened papers that had been selected for further processing. We have shown the number of primary selected studies in Figure 3.
As it is illustrated, there is an increasing tendency in applying intelligent speed assistance in transportation systems. We expect to observe more research studies providing speed advisories in CV environment, in contrast to the significantly reduced number of articles in 2020 as compared to 2019.

Significant keyword distribution
As we mentioned in Section 2.1, we used several vehicle speedrelated keywords to retrieve the selected primary studies. In this regard, Figure 4 shows the popularity and appropriate search keys, which are more common for motorised and ground vehicles research directions.

ANALYSIS OF VEHICLE SPEED ASSISTANCE APPROACHES IN ITS
This section analyses vehicle speed assistance approaches based on the three primary perspectives or aspects, namely high-level objectives, driving goal(s), and technology. Figure 5 depicts the details of the presented taxonomy tree. Considering the high-level objectives perspective, each primary research paper was studied read comprehensively, and related qualitative and  quantitative data were analysed and described thoroughly as seen in Table 5.

Vehicle speed assistance perspectives
By reviewing the related literature, on the one hand, we can observe that the speed assistance methods in ITS have several high-level objectives to deal with such as GLOSA, intersection conflict, and traffic harmonisation as illustrated on right side of Figure 5. On the other hand, it pursues to achieve different goals, namely eco-driving, safety etc. They also use different technologies in order to implement the system, which we call technology perspective in the presented taxonomy.

Driving goals perspective
In this section, we concentrate on goals/outcomes that any particular vehicle speed assistance method provides to individual drivers depending on the high-level objectives.
• Safety: Safety is one of the primary driving goals, which is the prominent concern of individual drivers. As the literature indicates, excessive and inappropriate speed remains a crucial factor in fatal crashes, and all countries report that speed contributes to fatal road accidents [13]. In this regard, we present several safety indicators that play a crucial role in the reviewed vehicle speed assistance systems.
1. Minimum safe distance: The minimum safe distance is the required longitudinal or lateral distance that prevents rear-end and side-angle crashes caused by sudden speed changes.

Time to collision (TTC): TTC captures the time that
remains until a collision between two preceding vehicles would have happened [14]. 3. Road friction level: Unawareness of road surface friction level is considered as one of the leading causes of vehicle rollover and safety loss [15]. Vehicle speed assistance systems usually take advantage of estimating friction level before recommending appropriate speed to drivers [16]. • Eco-driving: Due to significant gas emissions, in particularly CO2, in the transportation sector, researchers are looking for ways to solve this problem by reducing fuel consumption and emissions. Eco-driving and green driving are driving techniques and approaches that can lead to less pollution and energy loss. • Time-saving: One of the crucial issues in transportation systems is reducing total traveling time and road congestion. In recent years, many researchers have proposed adaptive traffic controllers in order to reduce vehicles' total travel time. It is also worth mentioning that choosing/suggesting optimal speed by/to drivers for different goals such as minimising travel time has been of great interest to many researchers. • Passenger comfort: Passenger comfort refers to the level of individual occupant' sick feeling due to driving speed. If the driver can maintain an appropriate speed when approaching a curvy section of the road or the preceding vehicles, the passengers will feel more comfortable. The speed assistance system seems to be able to help the drivers to choose the appropriate speed level in roads.

Technology perspective
The technology perspective of vehicle speed assistance systems includes both vehicular communications [17] and vehicle automation levels [18]. As Table 5 shows, the studied literature reveals a variety of technologies developed to provide appropriate speed to vehicles.

High-level objectives perspective
There are several high-level objectives that vehicle speed assistance systems try to achieve. For example, concerning the optimising the traffic flow at signalised intersections using CV technology for vehicular communications, GLOSA is one application that uses timely and accurate traffic information to advise an appropriate speed to drivers, which allows them to pass the intersection by fewer number of stops. Furthermore, many recent vehicle speed assistance systems aim to harmonise traffic and smoothen traffic flow at highways. In the following section, we will provide a comprehensive analysis of recent research according to the presented taxonomy.

Green light speed advisory
Aiming at an Eco-driving strategy to reduce fuel consumption and emission, Stebbins et al. have considered full individual vehicles' trajectories to provide acceleration advice continuously until passing a signalised intersection [19]. The authors proposed several acceleration advice algorithms to avoid red light running crash by taking minimum safe distance to intersection into account. They have formulated the problem as an optimisation problem and have provided a self-developed webbased simulator to analyse the effectiveness of the proposed approach according to the average number of stops in a carfollowing model. Several studies [20,21] have considered a multi-objective energy-travel time optimisation problem to provide real-time speed advice. In [21], Simchon et al. proposed a dynamic speed advice algorithm which optimises the number of stops before a traffic light, considering time and energy penalties. The simulation results of this study have shown a near-optimal solution compared to some existing methods but with a significant calculation time, which can be extended to all types of vehicles.
Considering hybrid electric vehicle (HEV), Luo et al. have proposed a speed advisory strategy to overcome high fuel consumption and passing-time for continuous intersections [22]. They have assumed that the HEVs have access to all traffic lights' schedules, which continuously help them find an optimised speed. The proposed non-linear optimisation function was solved by a genetic algorithm (GA), and simulations have shown fuel consumption and travel time improvement. The optimisation function was constrained by maximum safe acceleration.
Qiu et al. have modelled decentralised global energy management of HEVs and proposed a sequential quadratic solution to solve the optimisation problem [23]. They have considered a safety measure for provided speed guidance strategy focusing on travel time and fuel consumption.
Yu et al. have considered the presence of mixed traffic and CV accessibility to signal timing information, and Vehicle to Everything (V2X) covered traffic conditions to present a consensus-based speed advisory platoon model [24]. They aim to provide safety, Eco-driving, and travel time reduction by suggesting optimal speed to platoon leaders to efficiently pass the intersection. Furthermore, the platoon leader's speed trajectory plan is based on trigonometric-curve and logistic curves, which guide the platoon to depart from signalised intersections. The proposed method was solved by a consensus optimisation solution and safety criterion of car-following model was evaluated to show appropriate inter-vehicle safe distances.
Xu et al. have presented an intersection speed guidance model based on quasi-moving block theory, which takes ecodriving, safe distance, and travel time minimisation into account [25]. The proposed method considered platoon speed guidance consisting of both human-driven vehicle (HDV) and CV drivers and was evaluated using the ESTINET simulation model.
With the consideration of mixed traffic of electric vehicle (EV) and HDVs, Liu et al. [26] have proposed an eco-speed guidance method based on driver bounded rationality, which was studied in [27]. The numerical test of the proposed speed guidance model has shown acceptable performance in reducing emissions under different scenarios. Additionally, The authors investigated the impact on guidance performance concerning several of EV's market penetration rate (MPR).
In another study, Tang et al. have studied speed guidance method with combination of a car-following driving model in a single-lane road to reduce idle time during red lights, which results in a more safe and eco-friendly environment [28]. They have proposed a speed guidance strategy for each driver and divided the process into four steps as follows: search the leading vehicle in each cycle, search the feasible trajectory for the leading vehicle by using a proposed pruning algorithm, adjust each vehicle's expected speed in the fleet, and select the optimal trajectory for each vehicle in the fleet. They have shown that four cases in which the leading vehicle's speed in each fleet is adjusted based on real-time traffic situations. Finally, the authors have applied the proposed model to analyse the speed guidance effects on the car-following headway distance and fuel consumption during the process that it runs across multiple signalised intersections.   Yang et al. have proposed a traffic controller system in order to prevent rear-end crashes at signalised intersections by harmonising CVs' speed. The authors have utilised CV data and traffic sensor information to address safety and mobility problem in mixed traffic scenarios [29]. The proposed system has several modules such as safety function and mobility function to optimise arterial traffic flow and signal plan in real-time. The effectiveness of safety was provided in reducing the potential number of side-angle and rear-end crashes.
In another study, Liang et al. utilise connected automated vehicle (CAVs) information to optimise traffic light timing plans and provide speed guidance to individual vehicles aiming to decrease passenger travel time by reducing the number of stops at an intersection [30]. They have considered mixed traffic of CAVs and HDVs that cooperatively follow the determined suggested speed to overcome the stop and go traffic issue before intersections. The authors proposed a heuristic algorithm based on a rolling horizon optimisation framework. This study's simulated results have shown that the 40% penetration rate of CAVs contributes to less travel times.
Wan et al. have presented a speed advisory method in CV environments with the presence of human drivers [31]. The proposed optimal solution works for pre-timed traffic signals, which considers minimum headway gap, ride comfort, Ecodriving, and travel time to provide recommended speed to an individual driver. The authors evaluated their work using the Paramics microscopic simulation.
In the same context as previous studies, Liu et al. have proposed two different speed guidance methods (single vehicle and cooperative) considering CVs, which optimise the acceleration of vehicles approaching isolated signalised intersections to overcome travel delay with taking minimum safe distance into account [32]. They have also developed a simulation framework to analyse the effectiveness of proposed approaches.
In one recent study in the context of path-planning, Typaldos et al. have proposed an analytical solution to the non-linear optimal control problem, which was then applied to the GLOSA scenario to reduce passenger cars' fuel consumption [33]. The authors considered car acceleration as an input of the control problem and then solved the problem analytically using Pontryagin's minimum principle (PMP) approach. The presented solution has shown acceptable computation time and lower fuel consumption compared to most existing heuristic approaches.
Furthermore, in the same context with the consideration of platoons, The authors in [34,35] have proposed an optimal speed guidance method for vehicles passing signalised intersections. In [35], Feng et al. proposed a method based on platoon leader acceleration with composing of follower vehicle's safe longitudinal distance by using a reinforcement learning algorithm to increase intersection throughput. The simulation results confirmed the composition approach's effectiveness on fuel minimisation and the number of stops. In both of the research papers, acceleration and deceleration rates were used as the optimisation objectives [35]. Wu et al. have considered various platoons in advising appropriate speed in successive traffic signals passing to optimise the number of stops [36].
Focusing on plug-in hybrid electric vehicles (PHEV), Qi et al. have proposed a co-optimisation framework that provides a real-time longitudinal velocity planning solution based on all trip factors that influence fuel consumption [37]. The developed system inputs are signal planning and timing (SPaT) information, traffic condition, Powertrain operation, and speed trajectory data using vehicle to vehicle (V2V) / vehicle to infrastructure (V2I) communication. To test the proposed system. The authors have evaluated a field test using GlidePath prototype application, and the results have shown acceptable fuel efficiency at signalised intersections.
Due to the presence of queues at signalised intersections, He et al. have proposed a real-time speed advisory system for V2Ienabled passenger cars using high-resolution arterial traffic and SMART-signal system data [38]. The presented optimisation method was modelled as a multi-stage optimal control problem considering GHG emission model to lower vehicle fuel consumption. The authors have divided the optimisation problem into sub-optimal problems, which are solved by a meta-heuristic algorithm for more efficient computation.
Aiming to reduce fuel consumption by taking minimum safe distance and travel time constraints, Ye et al. have presented a meta-heuristic algorithm solution to an optimisation problem to provide speed guidance strategy of vehicle platoons at intersections [39]. The authors evaluated the proposed model using a simulation model of the intersection based on field data using VISSIM. Eventually, their numerical results have shown the algorithms' effectiveness in the control of the fuelsaving effects.
To optimise signal timing plans simultaneously with the speed guidance strategy, Wu et al. have proposed a speed guidance model at single intersections in CV environments [40]. The authors have considered queuing delay and signal delay caused by traffic control to guidance appropriate speed to vehicles passing the intersection smoothly. The final results have shown that the proposed dynamic optimisation could significantly reduce vehicle delays and fuel consumption of CVs when the penetration rate is 100%.
In the research conducted by [41], the authors addressed the network-wide traffic light coordination using advisory speeds and traffic light offsets. They have formulated the bandwidth maximisation problem as an optimisation problem, which is then solved by linear programming. The proposed solution aims at reducing travel time and fuel consumption in an arterial road with continuous intersections.
To address the high fuel consumption problem in urban networks, researchers in [42,43] have utilised infrastructure to vehicle (I2V) and signal timing information to advise appropriate speed to vehicles approaching a series of signalised intersections. The authors have utilised the Dijkstra algorithm to divide the urban network optimal control problem into path optimal problems, which led to a convex optimisation problem that could be optimised with low-computational resources for real-time uses. Both articles increased the road throughput efficiently.
Zhang et al. have proposed a platoon splitting and speed guidance to a fleet of vehicles passing isolated signalised intersections [44]. They aim at traffic efficiency, safety, and ecodriving with passenger comfort in mind. They have proposed a multi-objective optimisation method and signal timing plan to decrease the whole platoon energy consumption consisting of EVs.
Considering urban street networks of intersections, [45] have proposed a linear programming solution to propose a speed guidance method in CV environments. The authors have utilised a model predictive control (MPC) approach to overcome the dynamic and stochastic nature of traffic, which resulted in the improvement of travel time reduction. [46] have formulated eco-driving as a data-driven chanceconstrained robust optimisation problem and provided a dynamic programming solution to solve the problem. The authors considered the uncertainty of traffic signal timing, speed limit, and minimum headway distance constraints. The final results have illustrated 40% less vehicle fuel loss and the same level of arrival time compared to a modified intelligent driver model (IDM).

Non-signalised intersection conflicts
To overcome the speed management issue at unsignalised intersections, Chen et al. have proposed a safe gap-based acceleration and deceleration guidance method to achieve safety and fuel economy objectives [47]. Their system uses dedicated short range communication (DSRC) protocol with full compliance of automated vehicles in the traffic network. The simulation results of the study have indicated that vehicles experienced lower fuel emissions without compromising safety.

Traffic harmonisation
Due to pervasive environmental sources of data in smart cities, Gallen et al. have proposed a speed limit advisory for individual vehicles considering weather and fog conditions [48]. A further study by [16] has presented a framework that combines in-vehicle and off-vehicle data to estimate real-time car status. The proposed methodology utilises V2X communication to gather weather data, route structure information, and in-vehicle diagnosis data to better understand the surrounding environment to which the car should adapt. Galanis et al. have shown a recommendation speed as a use case of their proposed approach. In their presented work, weather conditions for a specific route are estimated by a trained classification tree, and road surface index calculation is performed based on the road's status to prevent rollover caused by inappropriate vehicle speed.
Ghiasi et al. have considered a section of single-lane highway as a geometry model to provide speed harmonisation. The authors proposed an algorithm for speed harmonisation that addresses mixed traffic scenarios, including CVs, CAVs, and human drivers [49]. They have used real-time sensor data and V2X communication capabilities to adjust CAV's speed to smooth traffic flow. The authors have noted that the proposed algorithm reduced fuel loss, emission, and traffic delay imposed by highway bottlenecks. Inverse time-to-collision (iTTC) measurement has been utilised to investigate the improvement of the proposed approach in terms of safety.
Focusing on reducing emission or energy consumption of a vehicle fleet traveling extra-urban routes, Liu et al. have presented a consensus-based optimisation solution [50]. Having data privacy of vehicles in mind, authors have considered a speed advisory system that utilises V2X communication to recommend a common speed. Performance of the proposed algorithm was evaluated by SUMO microscopic simulation and hardware-in-the loop (HIL). The results have shown the convergence duration of recommended speeds in dynamic nature traffics and appropriate vehicle energy/fuel savings.
In addition to the previous study's findings, Griggs et al. have proposed two consensus algorithms named leaderless and leader-based speed advisory systems, respectively [51]. The proposed systems are distributed and were designed for multi-layer networks, and an obfuscated signal preserves the privacy of vehicles. It is worth mentioning that the leader-based speed advisory system considers a reference speed provided by an external entity for minimising a particular goal. The presented consensus optimisation problems were solved by the CVXPY python tool, and emulation was performed by SUMO with a combination of HIL.
Due to the importance of safety and effects of environmental factors for selecting appropriate speed by drivers, De Mello et al. have presented a fuzzy logic model to speed limit suggestions in highways [52]. In this study, investigating main safety factors was conducted by the Delphi method and further classified by a soft decision tree, and the proposed model was simulated in a single lane highway traffic scenario. They have considered road structure and weather information to estimate speed reduction, which decreases rollover possibility.
Ma et al. [7] have proposed a simple linear speed recommendation system considering the speed-space relationship influenced by the work carried out in [53,54]. They have tested the provided method experimentally in a traffic scenario consisting of remote traffic microwave sensors and V2I-equipped vehicles to measure the average speed. They have shown that the proposed speed harmonisation algorithm has decreases traffic congestion with no increase in travel times.
In the context of autonomous vehicles (AV) on congested highway zones, Malikopoulos et al. have aimed at providing speed control with the consideration of a safe, eco-friendly, and low delay driving strategy [55]. The authors formulated the problem as a control problem constrained by minimum safe distance and then provided an analytical solution to it. MAT-LAB and VISSIM simulation framework evaluated their proposed speed control method, and the results have depicted significant fuel consumption reduction and improvement in traffic mobility.
As there are several fuel economy contributors, such as dynamic traffic conditions, route geometry, and driving style, Ozatay et al. have proposed a dynamic programming (DP) solution to find the optimal velocity profile for a given route to decrease fuel consumption [56]. The authors have presented a cloud-based architecture that considers route information and vehicle dynamics in calculating a proper recommended speed. They have tested the proposed method in a real driving scenario by a car equipped with a visual interface and have resulted in 5-15% improvement in fuel economy.
In a more recent study, Abdelghaffar et al. have proposed a dynamic freeway speed controller to overcome vehicle emissions and crashes by decreasing the number of stops caused by traffic congestion [57]. They have presented the method based on sliding theory and CV environment. The study's simulation results in Los Angeles have illustrated the significant reduction in travel time and CO2 emissions by 12.17% and 2.6%, respectively.
Kamal et al. have proposed a comprehensive eco-friendly speed assistance system that utilises inter-vehicle communication and analyses environmental situation (e.g. a preceding vehicle, road slope) to recommend appropriate optimal speed to an individual vehicle [58].
To achieve an optimised speed profile concerning safety, travel time, and fuel consumption, [59] have proposed a dynamic programming solution for safe speed profile calculation based on action rules obtained offline considering a set of elementary road stretches and legal speed.
In another recent study, [60] have proposed a reinforcement learning algorithm for merge control at highways. In their research, Q-network learns how to harmonise speed to reduce fuel consumption and mitigate traffic congestion by controlling the speed of CAVs.
Huang et al. have proposed a hierarchy optimisation framework for CAVs [61]. The proposed framework includes real-time traffic estimation and global speed optimisation based on [62] for less fuel consumption and safe lane-change. The genetic algorithm used to solve the optimisation problem has shown sufficient convergence time for real-time usage. The simulation results in VISSIM have shown acceptable fuel consumption reduction.
Han et al. [63] have proposed an energy-efficient eco-driving control system for connected electric vehicles to adjust the speed optimally while keeping a minimum safe distance and obey a maximum speed limit to avoid a collision. They have formulated an optimal control problem subject to a safe distance and speed limit to minimise energy consumption. They have provided an analytical state-constrained solution to show its real-time usage. They have shown that the proposed system does not increase travel time and minimise energy consumption. To overcome the trade-off between mobility and fuel consumption, Weng et al. have presented a model-free approach for optimal speed adjustment to a platoon of ground vehicles [64]. The authors have proposed a Nelder-Mead gradientfree heuristic optimisation solution, which results in the desired balance between mobility and fuel economy. The proposed method was evaluated by a combination of simulation and a networked engine-in-the-loop.
Park et al. have developed a speed control strategy that avoids inter-vehicle crashes to improve passenger safety in a CAV environment [65]. The proposed methodology optimises constructed risk map in order to provide appropriate vehicle speed control. The developed algorithm was simulated using VIS-SIM simulation environment to assess the effectiveness of vehicle safety.

Others
To provide safe speed assistance in autonomous route following, Olivares et al. have proposed a cost-effective visionbased solution that utilises a single monocular camera [66]. The presented work provides automatic steering control and speed assistance for the passenger car that travels along a predefined path. The computer vision approach recognises a line painted on the road and recommends appropriate speed to the autonomous vehicle speed controller. To prove the proposed method, the authors have considered a real driving scenario and have shown the car could effectively travel a distance of 47 km.

DISCUSSION
This section presents the analytical classifications of primary studies included in this systematic literature review to answer all research questions provided in Table 1. Furthermore, we provide some analysis charts according to the technical questions to give a complete view of the studied literature. At the end of  Table 5. RQ1: What are the most recent driving goal(s) that vehicle speed assistance systems aim to achieve? Figure 6(A) shows the objectives which have been addressed the most by the vehicle speed assistance systems. As the figure depicts, the combination of safety, eco-driving, and travel time reduction was the most important issue that researchers have addressed.
RQ2: What are the high-level objectives of reviewed vehicle speed assistance techniques?
In order to answer RQ2, we have categorised different highlevel objectives based on traffic scenarios in Figure 7. It was concluded that the most recent speed assistance techniques have focused on advising appropriate speed to vehicles approaching intersections. Moreover, highway traffic harmonisation comes second in rank among the applications reviewed in this SLR.
RQ3: What methods are provided to solve the vehicle speed assistance issue? Figure 6(B) gives an insight into the applied solutions to overcome the vehicle speed assistance issue. As illustrated in this figure, the studied literature have mainly solved this problem using optimisation techniques and heuristic/metaheuristic algorithms.  Figure 8 shows the analysis environment in vehicle speed assistance studies in ITS. Simulation environments using microscopic simulation models such as VISSIM and MATLAB are the most popular simulation environments. It is also worth mentioning that several papers have analysed the system by a hybrid evaluation method and hardware-in-the-loop.
RQ5: What are the challenges and future issues in intelligent vehicle speed assistance?
To answer RQ5, we discuss the remaining limitations and gaps in the primary studies considering three mentioned perspectives, the speed assistance systems high-level objectives, driving goal(s) and technology.
First, from the vehicle speed assistance objective perspective, we divided the existing vehicle speed assistance systems based on driving scenario and high-level objective into four major groups: GLOSA, non-signalised intersection conflict, and traffic harmonisation. Several research papers such as [26,30,44] have focused on systems that provide appropriate speed to individual vehicles at single intersections. However, there might be a network of intersections in an urban scenario that must be taken into account. Furthermore, other research such as [19,35,49,55] have considered single lane scenarios where vehicles would not take lateral movement actions. Providing vehicle speed advisory with the consideration of lateral movement and overtake manoeuver is still an open problem and needs further research.
Second, according to some recent research focusing on the driving goals and outcomes perspectives of vehicle speed assistance systems such as [70,71], driver acceptance level or preferences is a factor that needs to be taken into account to improve ITS. As Table 5 illustrates, some research such as [30,45] have concentrated on only one of the driving goals reviewed in the literature. On the other hand, in certain other studies such as [49,50,55] the authors have presented speed advisory systems with a focus on several driving goals. It seems noteworthy that essential to any practical solution is its capacity for the inclusion of different driving goal preferences. Indeed, a vehicle speed assistance solution is desirable, provided that it can be flexible in the face of different driving goals in a platoon simultaneously.
Third, from the technology point of view in vehicle speed assistance systems, it is worth mentioning that the deployment of many of the proposed methods in the literature is only applicable to high-tech vehicles [35,65,66] and inter-vehicle communications. Given these limitations, there is a need to propose a speed recommendation/assistance model for legacy and non-connected vehicles as well. Based on the outcomes of this review and our observations, we provide the following research directions of vehicle speed assistance systems that are worth further research:

Special-infrastructure less approaches
Although connected and autonomous vehicles can communicate with each other and road-side infrastructures, to overcome the lack of these infrastructures in developing countries, providing solutions that use cellular networks and smartphones [72,73] for recommending appropriate speed to individual drivers needs further investigation.

Inclusion of driver preferences
By studying the literature, we have observed that addressing multiple goals or sub-goals is a crucial issue. There is a need to propose a speed recommendation model for legacy and nonconnected vehicles that considers safety and any other secondary goal (possibly conflicting) selected by individual drivers. Hence, providing speed recommendation services based on individual driver goals (e.g. safety and eco) can contribute to more compliance and driver-satisfactory with having safety as a primary driving goal.

Distributed and scalable algorithms
Since vehicle speed assistance and recommendation services require real-time computation [59], proposing lightweight/ distributed solutions for road-user safety could be another area of further research.

CONCLUSION
In the ITS context, intelligent speed advisory/assistance for onroad passenger vehicles has been a research subject for a long time. This research has analysed recent studies on intelligent speed assistance approaches that overcome adverse environmental and road safety problems. Furthermore, it is interesting to summarise these techniques and analysis of the outcomes and objectives achieved. By doing this research, an inclusive comprehension of the vehicle speed assistance systems, the open issues of this field, related challenges, and future directions were obtained. This study presented a systematic review of 355 papers published between 2011 and 2020. Eventually, this paper has analysed 50 papers that had passed quality assessment process focusing on vehicle speed assistance systems. The statistics number of research papers by year in this research field illustrates an increasing trend from 2014 to 2019, expecting more research to be carried in 2020-2021. According to the analytical discussions, the paper addressed five research questions. It was observed that 62% of speed assistance systems mostly focused on providing appropriate speed for vehicles passing single intersections, and traffic harmonisation on highways ranked second by 34%. Additionally, providing solutions to achieve a combination of driving goals, namely safety, Eco-driving, travel time, and comfort, is an ongoing issue. According to RQ4, it was discovered that 76% of the reviewed articles evaluated their proposed method using simulation tools, 16% implemented their proposed approach by field test, and 8% offered hybrid evaluations of the proposed methods. Eventually, in answering RQ5, we discussed the remaining limitations and gaps in the primary studies considering three mentioned perspectives, the speed assistance systems high-level objectives, driving goal(s), and technology. we have observed that there is a lack of driving-goal based speed recommendation systems for human-driven vehicles without the need for V2X communications and high-tech vehicles. Nevertheless, considering the growing number of research papers in this field, it is possible to conduct a further survey focusing on law-enforcement/management aspect of vehicle speed assistance systems.