CAV (Connected autonomous vehicles)

 

CAV (Connected autonomous vehicles)

Abstract

              Connected and autonomous vehicles (CAV) may be a transformative technology that has great potential to change our daily life. Therefore, CAV related research has been progressive expressively in recent years. This broadsheet does a whole review of five selected subjects that exist in the gluttons of CAV research: (i) inter-CAV communications; (ii) security of CAVs; (iii) connection control for CAVs; (iv)  collision-free steering of CAVs; and (v) pedestrian detection and protection. It is believed that these topics are essential to ensure the success of CAVs and need to be better understood

Introduction

              Connected and automated vehicles (CAVs) (a.k.a.connected) and autonomous vehicles and  (driver-less cars) is a transformative technology that has great potential for reducing traffic accidents, enhancing quality-of-life, and improving the efficiency of transportation systems. Bajpai (2016) showcased the positive effects CAVs should have, compared to present experiences. For example, the reduction of distance in amid vehicles for increased ability but not raising delay times ends with higher throughput. Other benefits may include declined emergency room patients, reduced car insurance premiums, and smaller sizes of traffic enforcement departments. Also, Bajpai noted that sharing a vehicle through third-party companies or individuals would lower the number of cars owned. The number of vehicles per household may drop with the availability of driver-less cars returning home to be used by other household members. Ride-sharing with autonomous vehicles can have a greater impact on available properties located within densely populated areas. Researchers (MR Cagney, 2017) pointed out that individuals who did not have a license to drive would beef it greatly from CAVs by relying less on traditional forms of public transportation. The requirement for large parking areas will go away from the rising use of autonomous car sharing that constantly scans for potential fares to pick-up. The constant hunting for riders will drive up the actual miles travel per vehicle compared to current standards. As mentioned, road incidents should be lower than present figures with the removal of human error in autonomous cars. It is estimated the traffic accidents will drop by 70% in 25 years

 

2. Inter-CAV communications

 Drivers navigate roadways and intersections by applying the information located on roadside signs, on radio stations, or other forms gathered before entering the vehicle for travel. The combination of all that information is necessary to make the correct choice when a situation arises. CAVs require a similar stream of data to institute a proper protocol for roadway events. The ways these vehicles get that information is slightly dissimilar to the human counterpart. For example, vehicle message allows for vehicles to communicate information between each other about various environmental vicissitudes like an outsized influx of traffic (Fang et al., 2017). The NHSTA forecasts that efficiently applying vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications could possibly reduce and/or remove up to 80% smashes of any kind from non-impairment (NHTSA, 2017). Fang et al. make note of two other forms of vehicle communication included in intelligent transportation systems: vehicle-to-pedestrian (V2P) and vehicle-to-network(V2N). All these communication mechanisms are collectively known as vehicle-to-everything (V2X)

3. Security of CAV

 CAVs, along with any other computing platform, are susceptible to two forms of attack, passive and active. Passive attacks read the information that is being transferred between a CAV and another communication point, e.g., RSU and a different CAV. This threat type has lower risk compared to active attacks where the results may end fatally. Active attacks may consist of spoofing incorrect data, resending a previous message to obtain validated system keys, message modification of relevant data, or denying of service that prevents data transfer on an affected server where data transference is vital.

 3.1. Passive attacks

    A CAV will program a message covering verified velocity, location, a pseudonym, i.e., VA or VB, for the car, and other information to alert other vehicles nearby for safety purposes. Attackers may eavesdrop location information and car pseudonyms for the initial selection of the target car. Verified data, similar to broadcast safety messages, is transferred to an RSU through a virtual machine (VM) (Yu et al., 2013). VMs are secure cloud connections to local RSUs and the CAV. The data from the VM gives the attacker a second observable data group. In Fig. 2, Kang et al. (2016) explained a method that obtaining unique and precise information from a roadside location unit (RLU) about a target car can observe a tie between VM and Va. As long as the VM is unrevised through the exchanges from one RLU to another or among different locations, the attacker is allowed to collect location data continually

 3.2. Active attacks

   At the roadway level, GPS signals are susceptible to malicious jamming attempts because their power levels are low after long-distance attenuation. Using a jamming signal as an active attack, the GPS can be overpowered and rendered ineffective. The danger comes in the inability to determine when and where to stop without the locating signal. In an attempt to reduce the effectiveness of jamming, Oz emir and AK soy(2017) combine adaptive null-steering with antenna arrays to determine a direction of origin for unwanted multiple jamming signals. Once the direction of the signals is estimated, a waveform needs to be produced for counteracting the jamming. Using an orthogonal array optimization method conceived by Taguchi (Oz emir and AK soy, 2017) the system compares an incoming jamming signal to a proposed broadcast cancellation signal using a fitness function

 

4. Traffic intersection navigation

       The requirement for traffic lights at intersection diminishes when CAVs take over the roadways. In this section, we discuss both “centralized” and “de-centralized” intersection control approaches. The difference between the two is that the infrastructure makes decisions in the former methods whereas the CAVs make decisions in the latter.

4.1. Centralized navigation

    Li and Wang(2006) proposes a centralized cooperative driving mechanism for an isolated blind crossing. Their goal is to achieve both safety and efficiency using inter-vehicle communications, which is done by dividing vehicles within a virtual circle cantered at the junction into groups of maximizing size 3. To ensure collision-free driving, the authors come up with the concept of safe driving patterns, which are essentially vehicle pairs that can pass the intersection simultaneously without the risk of colliding into each other. A driving schedule can then be represented as an ordered series of safe patterns. The authors develop the basic solution tree generation algorithm to generate the possible driving schedules. By using the first-in-first-out property of traffic flow, the modified solution tree generation algorithm is proposed byte authors to prune the solution tree. Furthermore, the authors use the safety pair labeling algorithm to prune the unsafe driving schedules from the solution tree. To come up with the best cooperative driving schedule and vehicle trajectory, the authors divide the driving scenarios into four cases and derive an optimization problem for each case. Finally, the best driving plan is obtained by comparing all feasible solutions and finding the one that yields the least amount of time. The method proposed in the paper is verified by simulation. The authors discuss several future directions in the paper, including combating the complexity of the centralized approach, improving the tree-based algorithms for multiple-lane crossings cases, incorporating real-world driving scenarios, and relaxing some of the assumptions such as knowledge of perfect vehicle speed and position.

4.2. Decentralized navigation

     An earlier effort of decentralized navigation can be found in the study of Milanes et al. (2010), where the authors use a CAV and a manually driven car to do intersection control. Both cars are mass-produced vehicles, equipped with a differential global positioning system (DGPS), PCs, and Wi-Fi adapters. The major difference between the two vehicles is that the former's actuators, i.e., steering wheel, throttle, and brake, are modified for autonomous driving. The goal of the works to control the CAV based on the information of the other manually driven car to let the two vehicles pass an intersection collision-free. The control system contains two parts. In the first part, the CAV uses V2V communications to detect the other vehicle's location, velocity, and direction when it is within 80 m; the point of intersection is then estimated based on both cars' trajectories. In the second part, a rule-based fuzzy controller converts the input fuzzy variables to actuator control via a three-step process: fuzzification, inference engine, and defuzzification. Experimentation results show that based on the real-time V2V data, the CAV can either yield to the manually driven car or traverse the crossroad before it while ensuring safety

5. Collision avoidance

5.1. Maneuverability

    One of the key aspects to the success of CAVs is the ability to avoid collisions with other vehicles. The primary option for avoiding most accidents is braking. However, in some scenarios, braking is not the best option or even a viable one due to high levels of traffic or the need to accelerate. A secondary method that needs to be considered is steering. Whether it is steering, braking, a combination of both, or some other means, a CAV needs to be able to respond to many circumstances where an accident may occur with the best countermeasure. This is where vehicle maneuverability plays a key role in the success of CAVs.

 

5.2 vehicle networking

       Current forms of an accident, avoidance relies on the driver, sensors, and the vehicle's collision avoidance system. Nevertheless, the autonomous vehicles possess the potential to gain aid from other vehicles to avoid an accident. If a vehicle has no room to avoid a collision, the accident bound vehicle could potentially relay to another surrounding vehicle to adjust its speed and position. This will give more leeway towards the vehicle avoiding the accident. In general, through networking and estimation, autonomous vehicles can be systematized so that vehicles on the same network can assist one another in preventing collisions. Present studies have successfully run simulations, which test the effectiveness of networked autonomous systems with and without estimation (MoradiPari et al., 2016)

5.3. Control confliction

        For collision avoidance to truly be successful in autonomous vehicles, it must be fully integrated with the vehicle's trajectory, counter-sway, and mapping systems; in addition, it should take precedence over them whenever needed. Countermeasures must be programmed in case of a potential conflict between another control system and the anti-collision system. For example, model predictive control (MPC) can be used to generate the essential control loop to allow the system to choose collision prevention over all other factors (Funk et al., 2017). Specifically, Funke et al. utilize Reline et al.'s work (Redline et al., 2013), involving vehicle path following and collision avoidance, to design their vehicle dynamics criteria and controller. They were able to create and run tests where an autonomous cart drove around a track with a designated trajectory. They then proceeded to add a pop-up obstacle in the path of the vehicle, resulting in a successful change in its direction to avoid it. The tests provide good scenarios to be applied to accident avoidance of autonomous vehicle and can provide the framework for more complex and dynamic collision scenarios, e.g. lane changing

6. Pedestrian detection

       The center for disease control and prevention discloses that over 180,000 non-fatal injuries were due to vehicles in 2015. Also, reports from the National Centre for Statistics and Analysis revealed that there was a 9.5% increase in pedestrian-vehicle fatalities from 2014 to 2015 (NHTSA, 2017). Pedestrians pose the biggest hurdle to the success of CAS due to their high risk of injury and fatality when involved in vehicular accidents. CAS has been proven to aid in the performance of human drivers’ ability to reduce pedestrian-vehicle collisions. With the continuous advances in autonomous collision avoidance and due to the limits in human capabilities, CAS can prove to be safer for pedestrians than human drivers altogether. As can be seen in the collision avoidance section, braking is one of the principal autonomous accident prevention methods. Because AEB reacts 2 to 4 times faster than a human driver, the European New Car Assessment Program (NCAP) is implementing requirements that vehicles must have some form of AEB to receive high safety reviews. Currently, the primary AEB sensors being utilized to detect pedestrians are cameras, with radar sensors used as auxiliary aids. Cameras and short-range radars typically have the capabilities to classify walkers up to 60 m away, while long-term radars can go up to 180 m. Typically, AEB applies three warning zone that uses an additional measure in each zone as the pedestrian gets closer, from a visible and audible warning to full brake activation. However, a key problematic with present AEB systems is that they only take into account the longitudinal drive of pedestrians

7. Conclusions

       Within the past decade, the advancement of technology that is built into or applies to CAVs are improving at a significant pace. While recent studies and experiments that we have discussed show success, the next step will be to integrate the components into a single platform without security or operating conflicts. Combining collision and pedestrian avoidance into the intersection navigation control without elevated risk will be a major hurdle for the general population's safety. The opportunities for incidents reside in the urban intersection with large amounts of pedestrian and cycling traffic along with human-controlled vehicles and CAVs. It is imperative that each component needs to work flawlessly without slowing down or restricting the overall throughput of intersections and general roadways. Inter-CAV communication is the gateway that RSUs, intersection managers, CAVs, and other safety devices used to alert each other and pedestrians. Of the topics that have been discussed, collision and pedestrian avoidance are near the apex of importance in the need for preservation of human life.