1. Introduction
Road transport system is significant in the transportation of goods and people from one place to another via the road network. The transportation of goods and people relies on the distance, weight, volume, and type of cargo, as well as the local road infrastructure. The demand for travel within an area is met by the availability of transportation services, collectively referred to as transportation systems. Analysing the relationship between road capacity and traffic volume is crucial in transportation studies. However, there have been significant issues with road transportation systems, such as traffic gridlock, road accidents, and inadequate road infrastructure, especially in a connected and automated driving environment (Alex et al., 2023; Barberi et al., 2022; Pappalardo et al., 2022).
The movement of people from one place to another is the core framework of the transportation planning and decision-making of urban planners in metropolitan cities. These cities primarily focus on walking, cycling (Zhang et al., 2023), and public transportation, which are starting to depend on electric vehicles and other intelligent shared mobility. These intelligent shared mobilities assists in accelerating the conversion of fuel-driven vehicles to electric vehicles and minimizes global warming and air quality failures. Urban areas have created and developed mechanisms to encourage people to use electric vehicles and not fuel-driven cars and single-passenger public taxis to achieve this aim. This is to convince people that sustainable transportation will continue in the future. Recently, metropolitan cities have invested financially in road infrastructure and intelligent transportation systems, which are necessary to assist connected multimodal transportation systems. This comprises of shared autonomous electric vehicles to replace less effective bus lines (Alonso Raposo et al., 2019). In addition, previous research carried out by transportation and logistics researchers on autonomous vehicles stated that the aim is to minimize the dangers of vehicular collision (Glaser et al., 2010), and Gelbal et al. (2020) suggested a new approach to the prevention of collision with reference to low-speed autonomous shuttles comprising of pedestrians in a real-life traffic situation in comparison to other approaches. Their objective was to focus on the immediate environment and safety of their driving environment, such as steering or braking, without any human intervention.
The importance of introducing autonomous mobilities into the transportation of goods and services cannot be under-emphasized (González-gonzález et al., 2019; Stead and Vaddadi, 2019; Ukaegbu et al., 2021b). In recent years, researchers such as Severino et al. (2021), Stead and Vaddadi (2019) and Mattas et al. (2021) have argued that the demand for autonomous mobilities will play a significant role in mobility in urban areas and its effects on the transportation of goods and services coupled with the risks associated with this disruptive technology on road transportation. There may be a reduction in the production of carbon and travel time spent in the vehicle, which allows the driver to engage in some other things rather than driving all day. However, it is not impossible to reduce traffic congestion and vehicular accidents, which are majorly caused by excessive drinking of alcohol, drivers not adhering to traffic rules, and tiredness (Olayode et al., 2022a). The primary objectives of autonomous vehicles are illustrated in Fig. 1.
According to the previous studies (Fagnant and Kockelman, 2015; Kockelman et al., 2016; Litman, 2017; Olayode et al., 2020a, 2020b), autonomous vehicles are supposed to reduce accidents on the road and make travelling times shorter, accident-free, reduced excessive road maintenance (Bösch et al., 2018), ergonomically comfortable, and more viable (Anderson et al., 2014; Brown et al., 2014; Fagnant and Kockelman, 2014; Wadud et al., 2016) and reduce the costs associated with travelling. Autonomous vehicles can lead to a gradual reduction of the overall number of vehicles on the road depending on the real-life traffic scenario (Bösch et al., 2018; Chen et al., 2016; Zachariah et al., 2014; Zhang et al., 2015) and significant urban road capability (Brownell, 2013; Fernandes and Nunes, 2010; Friedrich, 2015). Using autonomous technologies to transport goods will revolutionize the logistics and supply chain industry and change the status of urban transportation to a more robust, intelligent, and innovative system. Drastically reducing the overall expenses of travel may likely lead to a significant increase in additional travel demand (Gucwa, 2014; Hills, 1996) and usher in a new wave of suburban urbanization and urban development (Glaeser and Kahn, 2004).
1.1. Research gaps and contributions
Many transportation researchers have published academic literature describing the technological advancement of autonomous vehicles and their application in road transportation (Denaro et al., 2014). However, there is a limitation in the literature review which outlines the negative disruption of autonomous vehicle technologies and the application of autonomous technologies such as autonomous vehicles in transporting goods and services. Another significant observation is that despite recent research carried out by Bansal and Kockelman (2017) and Litman (2017), in which they stated that in the next 20 years, autonomous vehicles would have been widely implemented in road transportation systems worldwide (i.e., level 4 or 5 of autonomous vehicles). However, to our best knowledge, there is no research in the academic literature that critically scrutinizes the historical concepts, applications, and technologies of autonomous vehicles.
The changes experienced by road transportation over the last two years remain unprecedented. The outbreak of COVID-19 and its variants have led to a renewed interest in autonomous vehicles and the application of autonomous technologies in the transportation of people, goods, and services. This paper aims to torchlight the importance of introducing autonomous technologies in the movement of goods and services, including their advantages and disadvantages. This literature review will provide a significant opportunity to advance our understanding of autonomous vehicles and make an important contribution to the field of connected and autonomous vehicles. Also, this literature review will explain the mode of autonomous vehicles comprehensively and systematically. The research gaps are summarized as follows.
-
•
Public perception and acceptance of autonomous vehicles. While several studies have explored public attitudes towards autonomous vehicles, more research is needed to understand how the public views this technology, what factors influence their perception, and how to enhance public acceptability.
-
•
Safety implications of autonomous vehicles. While there has been some research on the safety benefits of autonomous vehicles, more research is needed on the hazards and obstacles of integrating them into the road transportation system.
-
•
Legal and regulatory issues. To deploy autonomous vehicles safely and effectively, liability, data privacy, and cybersecurity must be addressed.
-
•
Integration with public transportation systems. To enable seamless and efficient passenger mobility, autonomous vehicles need to be integrated with public transit networks.
-
•
Human factors. Drivers, passengers, and other road users must be studied to ensure safe and effective autonomous vehicle operation.
The contributions of this research to the field of road transportation, especially in autonomous vehicles, are in threefold.
-
•
This research investigates the overall impact of autonomous vehicles on road transportation systems.
-
•
This research examines the emerging role of autonomous mobility in the context of the transportation of goods and services.
-
•
Drawing upon previous related studies, this research gives a comprehensive overview of the history and definition of autonomous vehicles, not excluding the public perceptions of different countries on using autonomous vehicles in their road transportation networks.
The overall structure of this literature review takes the form of eight sections. Section one gives a brief introduction comprising of the aim and contributions of this research. Section two gives an overview of the recent history of autonomous vehicles, and the third section is concerned with the methodologies used for this literature review. The fourth section presents autonomous vehicle technologies, including their advantages, disadvantages, and limitations. The fifth section of this research discusses the human public perceptions of autonomous vehicle technologies. The sixth section presents the theoretical dimensions of autonomous mobilities in the transportation of goods and services, including autonomous mobility-as-a-service (MaaS). The seventh section explains in detail the impacts of autonomous vehicles in road transportation systems regarding safety, traffic congestion and travel behaviour. Finally, the conclusion summarizes the implications of the findings and future research on autonomous vehicles.
2. A short history of autonomous vehicles
Autonomous vehicles were firstly envisioned in the early 19th century (Pendleton et al., 2017). The US, Germany, France, and Japan had R&D programmes from 1964 through the early 2000s to develop autonomous bus and truck platoons, intelligent vehicle systems, and video-based vehicle driving scene processors (Shladover, 2018). Automotive makers like Volvo have used autonomous driving technology since 2006 and introduced a fully autonomous test car (SAE levels 1 and 2) onto the road transportation network in 2017 (Shladover, 2018). In 2009, Google and other technology companies developed an autonomous vehicle within the SAE levels 1 and 2. By late 2020, WAYMO, a subsidiary of Alphabet Inc., had debuted its commercial AV prototypes, which had driven over 3 million miles in four US states. Since 2014, TESLA, another tech behemoth, has built electric vehicles with driverless capacity of operating 90% of the running time without human intervention (Shladover, 2018).
Completely self-driving vehicles, such as an NHTSA-level 4 autonomous vehicle (Aldana, 2013), ensured a rudimentary advance in car mobility. According to the National Highway Traffic Safety Administration (NHTSA) (2016), autonomous vehicles are expected to reduce road accidents and make travel times shorter and accident-free, less expensive (Boesch et al., 2016), more comfortable, and more viable (Kockelman et al., 2016), and substantially reduce the costs associated with travel. Children, the disabled, and the elderly will find travelling easier with AV (Anderson et al., 2014; Chong et al., 2013; Chun and Lee, 2015; Kockelman et al., 2016). Depending on the real-life traffic condition, autonomous cars can steadily reduce the number of vehicles on the road (Bösch et al., 2018) and substantial urban road capability (Brownell, 2013; Tientrakool et al., 2011). Human errors on the road will be reduced by autonomous vehicles, and this is projected to result in considerable improvements in human and vehicular safety, vehicular mobility, and road transportation sustainability (Olayode et al., 2020a).
Nowadays, road transportation is experiencing a worrisome issue regarding the implementation of an autonomous vehicles in road transportation; issues such as traffic safety and road visibility issues are emerging. These issues must be addressed before AVs become a reality (Easa et al., 2020). Assume that all of the futuristic predictions regarding self-driving cars come true. In that case, it will alter worldwide road transportation networks and transform urban transportation into a more robust, intelligent, and inventive system. Reduced travel expenses are likely to increase additional demand for travel (Gucwa, 2014), as well as a new wave of suburban urbanisation and urban development (Glaeser and Kahn, 2004).
Moreover, there have been some significant issues with autonomous vehicles. These issues, for example, are the autonomous vehicle's reaction time to environmental changes and low public confidence are their key issues. Autonomous vehicles are often questioned, such as “How soon will autonomous vehicles be part of our daily lives?”, “Is it reliable?”, “Will it decrease traffic and road accidents?”
2.1. Different levels of SAE automation
According to the conceptual framework of an autonomous vehicle, we have four levels of taxonomy in automated vehicles; these levels were developed in 2013 by NHTSA (Wadud et al., 2016). In 2014, a level of taxonomy in automation, as shown in Fig. 2, was developed by the Society of Automotive Engineers International (SAE). It was later upgraded in 2016 (Coppola and Morisio, 2016; Milakis et al., 2017; SAE, 2016a, b; Snyder, 2016). Later in 2016, NHTSA adopted SAE's levels of taxonomy and automation (NHTSA, 2016).
Fig. 2In conceptual theory, an automated vehicle system can be called an “autonomous” system if the automated system inside the vehicle can do the dynamic vehicle driving functionalities in a driving environment. Regarding the Federal Automated Vehicles Policy of the United States Department of Transportation, a vehicle is considered autonomous if it possesses levels 3–5 automated systems (Dot, 2016). The caveat about these autonomy levels is that they are not adequately maintained, and researchers have grouped all these levels of autonomy as autonomous (Shladover, 2018).
Driving a vehicle on the road requires essential functions such as localization, perception, planning, control, and management (Coppola and Morisio, 2016). Acquiring information about the immediate driving environment of an AV is significant to localization and perception. The availability of all these functions in a vehicle is what makes a vehicle autonomous. Suppose any autonomous vehicles have to communicate with various types of infrastructures to acquire information about their driving environment or negotiate their driving manoeuvres. In that case, it is referred to as a connected autonomous vehicle (CAV) (Shladover, 2018); however, when any human-driven vehicle, either manual or automated, has to communicate with different types of infrastructures to possess information, it is known as a connected vehicle (CV) (Coppola and Morisio, 2016; Hendrickson et al., 2014). Therefore, CV technology is complementary or synergistic to implementing autonomous vehicles to a certain degree (Shladover, 2018), even though connectivity is not a mandatory characteristic of an autonomous vehicle (Hendrickson et al., 2014).
3. Methodology
3.1. Review design
This research utilizes a systematic literature review focussing on understanding the applications, impacts, and public perceptions of autonomous vehicles. Fig. 3 shows the five-step approach used in this research. To achieve the aim of this research, this research was designed to give a systematic review of autonomous vehicles taking into consideration the following.
-
•
Autonomous vehicle technology.
-
•
Travel behaviour and traffic congestions.
-
•
Understanding the capability, impact planning, and policies of autonomous vehicles.
Fig. 33.2. Source of literature reviews
The search was only conducted on peer-reviewed manuscripts. The search was conducted from January 2021 to August 2021, and we only focused on articles published in the last ten years (2011–2021). Different types of searches were carried out by combining multiple keywords. The keywords applied were grouped into two significant categories. The first parentheses used in the first group were associated with the title of the journal and conference articles, and the second group was applied to the abstract of this research. The resultant search keywords were first verified by going through the abstract and reading the full text to verify their scope against the research objectives.
This research used three academic databases (i.e., Scopus, EBSCO and Web of Science) to search articles related to “autonomous vehicles” or “automated vehicle technology” or “driverless vehicle” or “self-driving vehicle” or “autonomous driving”. These journal articles were then filtered by a criterion dependent on the inclusion and exclusion of relevant journal and conference articles to the aim and objectives of the research. According to the research by Mallett et al. (2012), the inclusion and exclusion criteria applied for screening of data gathered are applied to achieve the aim and filter the data collected so that the valuable data will not be mixed with useless ones. The inclusion and exclusion techniques principles were created and applied to filter journal articles, select common features, and choose the most relevant articles that were gathered. A typical example is a journal article that is 15 years old and cannot be used to compare to the one published two years ago. A clearly defined inclusion and exclusion criteria will assist in sorting data gathered (journal and conference articles) on a homogenous level. Table 1 shows the inclusion and exclusion criteria applied in this research.
Table 1. The inclusion and exclusion criterion.
| No. | Inclusion criterion | Exclusion criterion |
|---|---|---|
| 1 | High peer-reviewed articles | Articles that are not peer-reviewed |
| 2 | Original articles | Duplicate articles |
| 3 | Articles written in the English language | Articles not written in the English language |
| 4 | Possessing one or more search keywords in the abstract or keywords of the article. | Articles do not have any search keyword in their abstract or titles. |
It is important to note that these criteria are applied in sorting journal or conference articles according to their level of relevance to the research; this is important to ensure that the journal or conference articles used are appropriate for the research. An overall of 623 research articles were gathered using the keywords when searching academic databases such as Web of Science, Scopus, and EBSCO. The research narrowed its search to only English-language journals and conference articles because English is a universal language that is easy to speak and understand and is well-known in the academic community. Another significant criterion adhered to during this research is that the research only focuses on peer-reviewed journal or conference articles to achieve a high level of credible research contribution. The table below summarizes the process of identifying the articles that are not excluded from the research.
3.3. Sample and sampling techniques
The up-to-date search carried out on these academic databases was investigated by applying the search keyword of “autonomous vehicles” or “automated vehicle technology” or “driverless vehicle” or “self-driving vehicle” or “autonomous driving” to identify related research that aimed at Autonomous Vehicle capacities and abilities. These up-to-date searches came out with 623 research papers, which were reduced to 453 and 362 articles after deleting non-peer-reviewed and duplicated articles (this is shown in Table 2). It was further reduced to 158 after deleting non-English research articles. Fig. 4 illustrates the breakdown of how the articles used for this research were analysed considering the three academic databases used: Web of Science, Scopus, and EBSCO. Fig. 4 shows the PRISMA flowchart that was developed for the literature review. Fig. 5 shows the analysis of the search keywords in Fig. 5, “AV” is autonomous vehicles, “AVT” is automated vehicle technology, “DV” is driverless vehicle, “AD” is autonomous driving, “SDV” is self-driving vehicle.
Table 2. Compilation of the selection and inclusion of articles used in this research.
| Academic database | Search keyword | Selected article | Deleting non-peer-reviewed article | After deleting duplicated article | After deleting non-English article |
|---|---|---|---|---|---|
| Scopus | Autonomous vehicles | 58 | 36 | 26 | 13 |
| Automated vehicle technology | 47 | 32 | 27 | 15 | |
| Driverless vehicle | 50 | 40 | 32 | 15 | |
| Autonomous driving | 73 | 63 | 47 | 10 | |
| Self-driving vehicle | 62 | 41 | 39 | 13 | |
| Web of Science | Autonomous vehicles | 68 | 57 | 42 | 16 |
| Automated vehicle technology | 52 | 39 | 29 | 14 | |
| Driverless vehicle | 34 | 24 | 19 | 10 | |
| Autonomous driving | 56 | 39 | 30 | 10 | |
| Self-driving vehicle | 33 | 22 | 18 | 12 | |
| EBSCO | Autonomous vehicles | 22 | 19 | 15 | 12 |
| Automated vehicle technology | 13 | 10 | 8 | 4 | |
| Driverless vehicle | 19 | 10 | 9 | 4 | |
| Autonomous driving | 26 | 16 | 16 | 5 | |
| Self-driving vehicle | 10 | 5 | 5 | 5 | |
| Total | 623 | 453 | 362 | 158 |
Fig. 4
Fig. 54. Autonomous vehicle technology
In the last few years, autonomous vehicle technology has been making headway in road transportation systems, especially in the shape of advanced driver assistance systems (ADAS), which can be found in research and public transportation vehicles. Autonomous vehicle technologies aim to reduce or increase vehicular crashes, enhance the mobility of people with disabilities, reduce greenhouse gas emissions, and encourage transportation infrastructure applications more effectively (Fagnant and Kockelman, 2015). One of the primary reasons why there is an acceleration of advancement of autonomous vehicle technologies is avoidable human-dependent errors such as human fatigue and distractions when driving. All these human errors have been known to be the primary cause of 93% of road accidents from a statistical survey from NHTSA (Fagnant and Kockelman, 2015). The processes involved in the navigation of autonomous vehicles are shown in Fig. 6.
Fig. 6Currently, a primary issue is the occurrence of a brand new, unsafe act of driving by drivers who do not want to adhere to or understand the guidelines and ethical obligations when it comes to using AV-related technologies (Kyriakidis et al., 2019; Lowry et al., 2015). Additionally, it is of paramount importance for self-driving features and vehicles to significantly enhance the driving safety of vehicles and pedestrians, improving the mobility and effectiveness of AV technologies. Pedestrians and drivers must have a rudimentary understanding of the abilities of the AV technology (Kyriakidis et al., 2015, 2019). This comprises important features like technological limitations and technology usage, not excluding the suitable occurrences to apply or depend on the AV technology.
It is important to note that the navigation framework of an autonomous vehicle can be categorized into five primary components (Fig. 6). According to Cheng (2011), these five components are perception, localization and mapping, path planning, decision making, and vehicle control. Perception applies sensors to regularly scan and supervise the driving environment, which is almost similar to the vision of a human being (Maurer et al., 2016). Other components such as localization and mapping algorithms can be used to evaluate the global and local environment of the ego-vehicle and map the environment by using the sensor data and perception outputs (Maurer et al., 2016). AVs' primary components such as path planning can be used to evaluate the likelihood of driving safe routes for the ego-vehicle depending on AV components such as perception, localization, and mapping information (Katrakazas et al., 2015).
The AV component functions as investigating the optimal driving route depending on the likely driving paths, the present state of the AV, and the information on the driving environment such as features of the road, traffic signs, and unpredictable weather conditions (Maurer et al., 2016). In addition to this, the module associated with the vehicle control will then evaluate the suitable vehicle command, such as the vehicular acceleration, the angle of the steering wheel, and the torque. The evaluation is done to adhere to the appropriate decision-making regarding changing lanes or manoeuvring (Gruyer et al., 2016). It is imperative to understand that the navigation process of an AV involves an elevated frequency level of a recursive process. This elevated frequency enables AVs to supervise and control movable objects efficiently and possess high speed, for example, pedestrians, motorbikes, and vehicles (Julier and Durrant-whyte, 2003).
It is a common occurrence that vehicles usually possess their drive-by-wire functionality. This statement means that various distinct driving mechanisms, including but not limited to steering, can be supervised and controlled by a joystick or sophisticated computer software. This offers an interface between the hardware and software. The sensors fitted in the vehicles are interconnected to a group of computing systems connected to share data. The computing systems comprise the output connected to the steering wheel's motors, brake, and throttle. This computing system drives the vehicle.
Autonomous vehicles' intelligence allows them to autonomously navigate a traffic flow at a freeway or a road intersection can be found in the vehicle's software. This intelligence software functions as a tool for mapping the sensory percept of the sensors' sensors to the actuators' activation. However, the mapping is complicated, considering that many sensors create so much data information applied in complex manoeuvres navigation. Therefore, a gradual and hierarchical method is adopted, at which point having a proper clarity of the function ability of sensor percept occurs in various ways applied to create a world map. Another use of a sensor in autonomous vehicles is in the determination of the location of the vehicle. The autonomous vehicle's motion arrangement depends on a map representation of the world and the vehicle's GPS. Motion planning occurs in various distinct stages of hierarchies. The trajectory of the autonomous vehicle created by motion planning is imperative to the control algorithms. The vehicle can be controlled and send necessary signals to the actuation units.
Autonomous driving is therefore in continuous evolution and development. Although current automotive innovation has reached astonishing levels of automation, an even greater acceleration is expected in the coming years. The ultimate goal, which many car manufacturers are aiming for, is to reach level 5, ensuring that the driver is always safe and comfortable to drive.The automotive industry is working on optimising technologies to be implemented in autonomous vehicles and improve road transport, logistics and personal vehicles. Sensors are certainly critical components of autonomous vehicles. They can detect objects in the path of the vehicle and act as eyes, providing data from which autonomous decisions can be made. Cameras play a key role in image classification and interpretation (texture) and are currently the most valuable and widely used sensor type. They are the main sensors used by advanced driver assistance systems (ADAS) for frontal vision and recognition of traffic signals. The data acquired by the sensors convey information of crucial importance for vehicle decision-making. Several of the sensors used in today's autonomous vehicles incorporate advanced functionalities that enable them to handle all possible driving scenarios in all environmental conditions.
High dynamic range (HDR) is a critical and critically important specification of camera sensors. This parameter defines the sensor's ability to record information in different lighting conditions within a scene (where obviously there will be lighter and darker parts) without over- or under-saturation. The use of sensors with higher and higher resolutions is another important technological trend in the field of autonomous vehicles. Autonomous vehicles must not only be able to see further, but also detect objects of any size. This makes it possible to increase the number of pixels per degree on the object and gives greater efficiency to the automatic driving algorithms, which are then able to perform object detection and classification tasks (Royo and Ballesta-garcia, 2019).
For autonomous vehicles of level 3 and above, the dependence on hardware and software for decision-making is very high, not to mention that the vehicles themselves are more exposed to the danger of cyber attacks and fraudulent access (hacking). This has led to the development of numerous protection and security features. Another technology used in autonomous vehicles is LiDAR, although the cost is high. LiDAR's ability to create a 3D point cloud sets it apart from other sensing technologies. The objective is not only to detect an object on the road, but also to make a classification, provide feedback to the decision-making system and modify the vehicle's route when necessary. The reflectivity of an object also plays a crucial role. While current LiDARs are able to remotely detect objects with high reflectivity, further improvements are needed to enable these devices to detect objects with low reflectivity over long distances (Royo and Ballesta-garcia, 2019).
Radar is another popular type of sensor that is of critical importance for the future of self-driving vehicles. The advantages of radar are evident during night-time driving and in all weather conditions. Millimetre-wave radars are a particular class of radars that use electromagnetic waves with short wavelengths and are rapidly gaining popularity in autonomous driving applications. By acquiring the reflected signal, a radar is able to determine the range, speed and angle of objects. A further advantage is that more integrated and compact designs can be achieved without the need for large, visible antennas (Bilik et al., 2019).
Autonomous vehicles are undoubtedly one of the most eagerly anticipated innovations that will radically change the lifestyles of millions of people around the world. As this technology becomes more accessible to the general public, the overall effects-whether positive or negative-will have a significant impact. In particular, traditional and new car manufacturers are pursuing several actions aimed at developing both personally owned production vehicles and robo-taxis.
Despite advances in technology and interest from the public and businesses, the autonomous vehicle industry is still facing many challenges, foremost of which is safety. Although with some progress on safety, the regulation that could enforce safety requirements is largely absent, and it is still unclear how local authorities will administer the laws and regulations governing the use of autonomous vehicles. Some states have enacted legislation for autonomous vehicles.
Another challenge with the technology is trying to understand the area in which AVs can be most useful. One industry sector that is crying out for autonomous vehicles is the truck sector, which suffers from a widespread shortage of drivers. An autonomous vehicle on the motorway is much easier than driving an autonomous vehicle in the city. Autonomous vehicles are not at the stage where car manufacturers thought they would be by now. However, some vehicles are operating commercially with safety drivers inside. While some are focussing on trucks on highways because it's easier than driving in cities and doesn't contain many intersections, others are limiting the scope of their version of autonomous vehicle technology by geography, location or weather conditions.
In general, autonomous vehicle technology is still firmly in the experimental phase. It is possible to divide the mission of an autonomous driving system into four main steps: The first step is detection and perception, using a software system on top of an array of sensors including cameras, radar, lidar and ultrasonic sensors. The sensors observe the world around the vehicle to detect its surroundings. The next step is to make sense of all the different objects surrounding the vehicle (e.g., object recognition, pedestrians, cyclists, other vehicles and roadside objects). After prediction, the next step is to plan a route for a vehicle to drive through that environment. The final step is control. This is where the system sends signals to the steering, brakes and accelerator to get the car or truck where it needs to go. One of the thorniest problems for autonomous vehicle technology is the prediction phase. While it might be easy for the autonomous vehicle to figure out what other vehicles might do, pedestrians are relatively unpredictable and can easily change their minds at any time.