Experimental validation of connected automated vehicle design among human-driven vehicles

https://doi.org/10.1016/j.trc.2018.04.005Get rights and content

Highlights

  • The V2X-based control of a connected automated vehicle among human-driven vehicles.

  • Connected cruise control (CCC) algorithms designed and tested.

  • Benefits of beyond-line-of-sight information on safety demonstrated experimentally.

  • Benefits of beyond-line-of-sight information on mitigating traffic waves demonstrated experimentally.

Abstract

In this paper, we present results regarding the experimental validation of connected automated vehicle design. In order for a connected automated vehicle to integrate well with human-dominated traffic, we propose a class of connected cruise control algorithms with feedback structure originated from human driving behavior. We test the connected cruise controllers using real vehicles under several driving scenarios while utilizing beyond-line-of-sight motion information obtained from neighboring human-driven vehicles via vehicle-to-everything (V2X) communication. We experimentally show that the design is robust against variations in human behavior as well as changes in the topology of the communication network. We demonstrate that both safety and energy efficiency can be significantly improved for the connected automated vehicle as well as for the neighboring human-driven vehicles and that the connected automated vehicle may bring additional societal benefits by mitigating traffic waves.

Introduction

The advances in automotive and infrastructure technologies have revolutionized the safety and efficiency of road transportation in the past few decades. While traffic accidents and congestion problems continue to exist on our roadways (NTSHA, 2016, Schrank et al., 2015), advancements in automated driving technologies promise a much safer and highly efficient future of transportation (Stern et al., 2018, Aeberhard et al., 2015, Mersky and Samaras, 2016). However, many automated vehicles today only use on-board sensors to perceive the environment, which may not be robust to various driving conditions. In particular, as on-board sensors are only able to obtain current information about the immediate neighborhood (Harding et al., 2014), these automated vehicles often have difficulties anticipating the motion of surrounding vehicles reliably. Therefore, many disengagement incidents happen when an automated vehicle has to interact with nearby human-driven vehicles (California DMV, 2016).

In order to facilitate the implementation of automated vehicles in real traffic, it is desirable to introduce beyond-line-of-sight information through vehicle-to-vehicle (V2V) or vehicle-to-infrastructure (V2I) communication (Montemerlo et al., 2015, Urmson et al., 2017). These are often grouped as vehicle-to-everything (V2X) where X may also incorporate other traffic participants like pedestrians or bicyclists. Previous studies have found that V2X communication may further amplify the benefits of driving automation (Vander Werf et al., 2002, Shladover et al., 2015, Talebpour and Mahmassani, 2016). A prominent application of V2X communication is cooperative adaptive cruise control (CACC) where a group of automated vehicles follow each other while coordinating their motion via V2X communication (van Arem et al., 2006, Shladover, 2007, Milanes et al., 2011, Shladover et al., 2012, Wang et al., 2014a, Wang et al., 2014b, Ploeg et al., 2014b, Zheng et al., 2017, Li et al., 2017). In particular, CACC has the potential to improve fuel economy and traffic throughput (di Bernardo et al., 2015, Milanes and Shladover, 2014, Zhou et al., 2017, Lioris et al., 2017). However, as the penetration rate of automated vehicles is low, it is rare to find several of them driving consecutively on road, which may severely limit the implementation of CACC in real traffic (Shladover et al., 2015).

Thus, it is desirable to consider connected automated vehicle design in partially automated traffic environments (Jiang et al., 2017, Luo et al., 2016). In particular, we proposed connected cruise control (CCC) that is able to utilize motion information from multiple human-driven vehicles ahead (Orosz, 2016, Ge et al., 2017). Despite being categorized as an advanced form of CACC (Shladover et al., 2015), CCC exhibits different challenges from most CACC implementations. For example, CACC implementations often assume a priori knowledge about the parameters of each vehicle in the platoon (van Nunen et al., 2012, Englund et al., 2016), while CCC needs to be robust against variations is human parameters (Ge and Orosz, 2016). Furthermore, while most CACC research assume fixed communication topology among vehicles (Ploeg et al., 2014a, Turri et al., 2017), CCC allows the connected automated vehicle to select which motion signals are incorporated into its controller based on ad hoc V2X communication (Zhang and Orosz, 2016, Ge and Orosz, 2014, Qin and Orosz, 2017). Such flexibility allows ad hoc connected vehicle systems to form consisting of connected automated vehicles and surrounding human-driven vehicles equipped with V2X devices. When designed appropriately, the connected automated vehicle may reduce velocity fluctuations significantly and thus improve the safety, fuel economy and lead to smooth traffic flow of human-dominated traffic systems (He and Orosz, 2017, Avedisov and Orosz, 2017, He et al., 2018).

However, in order to harvest the aforementioned theoretical benefits, connected cruise control needs to be validated in real-world scenarios. Experimental and simulation studies on CACC have brought into life the benefits of exchanging motion information among automated vehicles (van Nunen et al., 2012, Mersky and Samaras, 2016, Sepulcre et al., 2013). Yet it remains a question whether motion information from real human-driven vehicles can indeed benefit a connected automated vehicle. As human car-following behaviors are much more volatile than automated driving systems, such experimental validations are indispensable for connected cruise control (Orosz et al., 2017). Therefore, in this paper, we conduct experiments both on a closed track and on public roads to examine the benefits of utilizing motion information from human-driven vehicles ahead for a connected automated vehicle.

By using real vehicles we experimentally validate that the connected cruise control design can be robust against changes in the car-following behaviors of preceding vehicles and the topology of the V2X communication network. That is, a connected cruise controller can handle the rich behavior of real human drivers ahead; and it can still be used when some of the surrounding human-driven vehicles are not broadcasting. We also demonstrate through experiments on public roads that appropriately designed CCC algorithms can mitigate traffic waves traveling along chains of human-driven vehicles, and thus, improve the safety and energy efficiency of the connected automated vehicle as well as the neighboring human-driven vehicles. That is, for the first time, we demonstrate that a single connected automated vehicle, which utilizes beyond-line-of-sight information appropriately, may improve human-dominated traffic flow while being integral part of the flow. These results validate the feasibility of CCC implementation in real-world scenarios and the experiments also suggest future research directions in connected automated vehicle design.

The rest of this paper is organized as follows: in Section 2 we describe the human car-following behavior that can be used as a base for the design of the longitudinal control of connected automated vehicles; in Section 3 we design a class of longitudinal controllers for the connected automated vehicle used in the experiments; in Section 4 we describe the experimental setup and the functionalities; in Section 5 we present a set of experiments that demonstrate how a connected automated vehicle can improve safety using motion information from multiple preceding vehicles; in Section 6 we present a set of experiments demonstrating traffic wave mitigation and improvements of energy efficiency; finally in Section 7 we conclude our results and discuss future directions.

Section snippets

Human car-following model

In this section, we model the car-following behavior of human drivers, which will later be serve as a base for the control design of the connected automated vehicle. For simplicity, we consider two vehicles driving consecutively in a single lane; see Fig. 1(a).

Based on (Orosz, 2016, Ge et al., 2017), we describe the dynamics of the human-driven vehicle i asṡi(t)=vi(t),v̇i(t)=αh,iVi(hi(t-τi))-vi(t-τi)+βh,iW(vi+1(t-τi))-vi(t-τi).Here the dot stands for differentiation with respect to time t,si

Longitudinal controller design for the connected automated vehicle

In this section we present the longitudinal controller for the connected automated vehicle used in the experiments. In order to be easily implemented and readily accepted by the passengers and other road users, the longitudinal controller is designed to have similar feedback structure as in the human car-following model (1) and (3). We consider the scenario where a connected automated vehicle 0 (blue) receives motion information from several human-driven vehicles ahead; see Fig. 3(a). The blue

Experimental setup for connected automated vehicle design and evaluation

In this section we provide some technical details about the human-driven vehicles and the connected automated vehicle used in the experiments, and we introduce the experimental scenarios used for connected cruise control design and evaluation. All human-driven vehicles used in the experiments are production vehicles retrofitted with V2X devices that broadcast standard 10-Hz basic safety messages (BSM) under the dedicated short range communication (DSRC) protocol (FCC, 2016, SAE J2735, 2016). We

Avoiding safety–critical scenarios on a closed track

To demonstrate the capabilities of the connected cruise controller (7), (8) we use a closed track to reconstruct two scenarios that could push human-driven or automated vehicles without connectivity to the physical limits. We show that by using V2X information appropriately the connected automated vehicle can avoid the safety-critical states. The control parameters chosen through these experiments will also be utilized for tests carried out on public roads described in the next section.

Mitigating traffic waves on a public road

Based on the design established in the previous section, here we evaluate the performance of the controller (7), (8) in real-world scenarios where the connected automated vehicle is traveling on a straight road behind six human-driven vehicles and followed by another human-driven vehicle. The vehicle at the front leads the vehicle chain with a series of mild braking events, a profile chosen because such mild speed perturbations have been observed frequently in urban and highway traffic. Through

Conclusion and discussion

In this paper, we proposed a general framework for the longitudinal control of connected automated vehicles and experimentally tested the performance of such connected cruise controllers that utilize beyond-line-of-sight information on real vehicles. We demonstrated that by using V2X communication, a connected automated vehicle is aware of a preceding vehicle obstructed by the road geometry, and is thus able to avoid a severe braking maneuver. We also demonstrated that by using motion

Acknowledgment

This research was supported by the University of Michigan Mobility Transformation Center. The authors would like to thank Commsignia, Inc. for their technical support and acknowledge the help of Sándor Beregi, Zsuzsanna Dobránszky, Ádám Kiss, and Henrik Sykora during the experiments.

References (55)

  • M. Wang et al.

    Rolling horizon control framework for driver assistance systems. Part II: Cooperative sensing and cooperative control

    Transp. Res. Part C

    (2014)
  • L. Zhang et al.

    Hierarchical design of connected cruise control in the presence of information delays and uncertain vehicle dynamics

    IEEE Trans. Control Syst. Technol.

    (2018)
  • Y. Zhou et al.

    Rolling horizon stochastic optimal control strategy for ACC and CACC under uncertainty

    Transp. Res. Part C

    (2017)
  • M. Aeberhard et al.

    Experience, results and lessons learned from automated driving on Germany’s highways

    IEEE Intell. Transp. Syst. Mag.

    (2015)
  • S.S. Avedisov et al.

    Analysis of connected vehicle networks using network-based perturbation techniques

    Nonlinear Dyn.

    (2017)
  • Barber, P., Engelman, G., King, P., Richardson, M., 2009. Adaptive cruise control system and methodology, including...
  • California DMV, 2016. Autonomous Vehicle Disengagement Reports 2016....
  • M. di Bernardo et al.

    Distributed consensus strategy for platooning of vehicles in the presence of time varying heterogeneous communication delays

    IEEE Trans. Intell. Transp. Syst.

    (2015)
  • C. Englund et al.

    The grand cooperative driving challenge 2016: boosting the introduction of cooperative automated vehicles

    IEEE Wirel. Commun.

    (2016)
  • FCC, 2016. Dedicated Short Range Communications (DSRC) Service....
  • A. Festag

    Standards for vehicular communication – from IEEE 802.11p to 5G

    Elektrotechnik Informationstechnik

    (2015)
  • Ge, J.I., Orosz, G., 2016. Estimation of feedback gains and delays in connected vehicle systems. In: Proceedings of the...
  • J.I. Ge et al.

    Optimal control of connected vehicle systems with communication delay and driver reaction time

    IEEE Trans. Intell. Transp. Syst.

    (2017)
  • J.I. Ge et al.

    To delay or not to delay – stability of connected cruise control

  • Harding, J., Powell, G., Yoon, R., Fikentscher, J., Doyle, C., Sade, D., Lukuc, M., Simons, J., Wang, J., 2014....
  • C.R. He et al.

    Data-based fuel-economy optimization of connected automated trucks in traffic

  • C.R. He et al.

    Saving fuel using wireless vehicle-to-vehicle communication

  • Cited by (0)

    View full text