Connected Variable Speed Limits Control and Vehicle Acceleration Control to Resolve Moving Jams

The vision of intelligent vehicles traveling in road networks has prompted numerous concepts to control future traffic flow, one of which is the in-vehicle actuation of traffic control signals. The key of this concept is using intelligent vehicles as actuators for traffic control systems, replacing the traditional road-side systems. Under this concept, we design and test a control system that connects 5 a traffic controller with in-vehicle controllers with Vehicle-to-Infrastructure communications. The link-level traffic controller regulates traffic speeds through variable speed limits (VSL) gantries to resolve stop-and-go waves, while intelligent vehicles control accelerations through vehicle propulsion and brake systems to optimize their local situations. It is assumed that each intelligent vehicle receives VSL commands from the traffic controller and uses them as variable parameters for the 10 local vehicle controller. Feasibility and effectiveness of the connected control paradigm are tested in simulation on a two-lane freeway stretch with intelligent vehicles randomly distributed among human-driven vehicles. Simulation shows that the connected VSL and vehicle control system improves traffic efficiency and sustainability, i.e. total time spent in the network and average fuel consumption rate are reduced compared to (uncontrolled and controlled) scenarios with 100% hu15 man drivers and to uncontrolled scenarios with the same intelligent vehicle penetration rates. INTRODUCTION Information and communication technologies enable cooperative vehicle infrastructure systems, where intelligent vehicles (IVs) connect with each other via vehicle-to-vehicle (V2V) communication and with road infrastructure via vehicle-to-infrastructure (V2I) communication. Numerous control concepts have been proposed during the past two decades with regard to traffic control 5 systems with IVs, most of which pertain to a hierarchical structure where road-based traffic control systems are placed on top of vehicle control systems (1, 2, 3, 4, 5, 6). Based on the spatial and temporal scope of the considered control system, several control levels can be distinguished in the hierarchy, i.e. network, link, (platoon and) vehicle levels (1, 7). The works in (1, 7) provide a detailed description on the hierarchical architecture. In this paper, we focus on the connection 10 between link-level controller and vehicle-level controller via V2I communications. The potential of connected traffic control systems with IVs compared to the traditional traffic control systems is twofold. Firstly, with V2I communications, the traffic controller can get individual vehicle information from IV sensors to estimate the traffic state. The data from IVs are usually provided in a finer spatio-temporal resolution and better accuracy compared to road-based 15 sensors (8, 9, 10). The second potential is the better actuation of traffic control commands, where IVs are used as actuators of road-based traffic control systems (1, 4, 11). The control signals are transmitted to the IVs via V2I communication, and the signals are used as commands for vehiclelevel controllers, i.e. IVs are forced to execute the control commands from traffic control systems. Next to these opportunities, road operators face some new challenges, one of which is 20 that the resulting traffic flow characteristics may change substantially with IVs traveling in the network. In (7), it has been shown that the formation and propagation of moving jams (also called stop-and-go waves) with IVs are quite distinct to those with all human-driven vehicles. Hence, the question of whether the current active traffic management (ATM) measures based on human-driven vehicular flow phenomena still work well with IVs arises. The answer to this question is of great 25 importance for road operators to develop ATM systems to control future traffic flow IVs. This paper presents a proof-of-concept study for the connected link-level variable speed limits (VSL) control system with intelligent vehicles equipped with Car-Following Control (CFC) systems. The main objective is to examine the feasibility of the connected control paradigm and to identify the potential improvements in control effectiveness. A VSL control algorithm that is 30 dedicated to resolve moving jams is employed as the link-level controller, while a CFC algorithm that controls vehicle accelerations based on model predictive control approach is developed and employed by the vehicle-level controller. The two-level controllers are connected via Dedicated Short Range Communications (DSRC). The connected control system is tested near a freeway bottleneck where CFC vehicles are randomly distributed in the network. Simulation experiments 35 with different penetration rates of CFC vehicles in the network are conducted to test the change in flow characteristics and in control effectiveness. In the sequel, we first describe the design and operation of the connected traffic and vehicle control system. Then the experimental design is presented. After that, the flow characteristics with IVs in uncontrolled scenarios (without VSL controller) are briefly discussed, followed by the test 40 results of the scenarios with the connected control system. Conclusions and suggestions for future research directions are summarized in the end. Wang, Daamen, Hoogendoorn, and van Arem 4 Vehicle controller (e.g. car-following controller) Vehicle actuator Vehicle system On-board sensors Disturbance Actual vehicle movement Reference vehicle control signal Traffic system with many vehicles Disturbance (Other vehicles position, speed, etc.) Local traffic measurements Traffic control signals from road-based actuators via short range V2I Vehicle state and control information via V2I Traffic controller (e.g., VSL controller) Road-based sensors (Loop detectors, etc.) Road-based actuators (VSL gantries, traffic lights, etc.) Vehicle state and control information via V2V Information from other cooperative vehicles via V2V Disturbance (demand, driver-vehicle heterogeneity, etc.) Traffic control signals directly from traffic controller via long range V2I FIGURE 1 Schematic representation of the bi-level control problem. Dashed lines are not covered in this study. CONTROL DESIGN OF CONNECTED TRAFFIC AND VEHICLE CONTROL SYSTEM Figure 1 shows a schematic representation of the connected control concept, with arrows indicating the information flow. At the upper level, the traffic controller estimates the freeway traffic state of the road network based on measurements from road-based sensors and on information from IVs in the network via vehicle-to-infrastructure (V2I) communication, and generates traffic control 5 signals, such as VSL. The traffic control signals are transmitted to and executed by road-based actuators, such as VSL gantries. The traffic control signal can also be transmitted directly to IVs from traffic controllers or from road-based actuators. As the vehicles in the network move based on the local interactions and the traffic control signals, the freeway traffic state changes, and the traffic controller enters the next control cycle, which is typically in the order of minutes. 10 At the lower level, the vehicle controller, such as the CFC controller, estimates the local situation surrounding the vehicle based on the measurements from on-board sensors. The traffic control commands from the traffic controller are used by the vehicle controller to compute the reference control signals, such as accelerations. The reference signal of the vehicle control system is executed by the vehicle actuators. The control cycle of the vehicle controller is typically in the 15 order of less than one second. The state and control information of the vehicle controller may also be transmitted to the upper-level traffic controller via V2I communication, or to other vehicles via V2V communication, which is not the focus of this study. The remaining of this section presents the assumptions regarding the operation of the connected control system, the algorithms at link and vehicle levels and the controller implementation 20 in a microscopic simulation model. Wang, Daamen, Hoogendoorn, and van Arem 5 Assumptions of Integrated Control System The following assumptions are made for the operation of the connected VSL control with CFC system on a freeway stretch: • There are loop detectors every 250 meters along the freeway, collecting aggregate flow and speed data every 30 seconds. VSL gantries are positioned every 500 meters along the freeway. 5 Measurement errors from detectors are not considered in this study, but are addressed in another work (12). • The VSL controller uses data from loop detectors to estimate and to predict the state of the traffic system on the freeway. • The transmission of VSL signals from VSL gantries to CFC vehicles are via Dedicated 10 short range communication (DSRC) (11). The DSRC range is 200 meters and the DSRC delay is negligible compared to the VSL control cycle. • A CFC vehicle detects the gap and speed difference with respect to the predecessor solely based on its own on-board sensors, e.g. forward-looking radar. The CFC vehicle predicts the behavior of its predecessor and determines its optimal accelerations to minimize its objective 15 function. • CFC vehicles are assumed to fully comply with the speed limits, while human drivers do not fully comply. Under the speed limit of 60 km/h, human drivers are assumed to choose their desired speeds at 72 km/h, which is in accordance with field test results (13). VSL Control Algorithm: SPECIALIST 20 At link level, a VSL control algorithm, namely SPECIALIST (SPEed ControllIng ALgorIthm using Shockwave Theory) (13), is chosen for this case study. Compared to other VSL control strategies (14, 15), SPECIALIST algorithm is easy to tune and has been tested on a Dutch freeway (13). It is a feedforward controller and resolves moving jams based on shock wave theory (16). Shock wave theory states that the wave front between two states in the left figure of Figure 25 2 has the same slope as that of the line that connects the two states in the fundamental diagram in the right figure of Figure 2. The approach to resolve stop-and-go waves consists of four phases and starts with a shock wave as shown in Figure 2. Phase I. A shock wave (shown as state 2 in Figure 2) is detected on the freeway. We assume that the traffic state upstream (state 6) and downstream (state 1) of the shock wave is in free flow 30 which is generally the case in real traffic. For illustration simplicity, we assume state 1 and 6 has the same state and density value. The theory holds when the two states are different (13). Phase II. As soon as the shock wave is detected, the speed limits of (60 km/h) upstream of the shock wave are switched on. This leads to a state change in the speed-controlled area from state 6 to state 3, and to the boundary between areas 6 and 3. State 3 has the same density as state 35 6. However, the flow of state 3 is lower than that of state 6 due to the combination of the same density with a lower speed. The front between states 2 and 3 will propagate backwards with a lower speed than the front between states 1 and 2, and consequently the two fronts will intersect and the shock wave will be resolved after some time. At the upstream end of the speed-limited area traffic will flow into this area, with the speed 40 equaling the speed limit and with a density corresponding to the speed limit, typically significantly higher than the density of state 3. This forms state 4. The propagation direction of the front between states 6 and 4 depends on the flow of state 4. Phase III. When the shock wave (area 2) is resolved, there remains an area with the speed Wang, Daamen, Hoogendoorn, and van Arem 6 0 100 200