Cooperative Automated Driving for Bottleneck Scenarios in Mixed Traffic

TL;DR

The paper tackles bottleneck scenarios in dense mixed traffic where benefits from connected automated vehicles (CAVs) emerge only with sufficient penetration. It proposes an emergent, V2V-based negotiation scheme among CAVs on the free and blocked lanes, using lightweight messages and two variants (counting and non-counting) to achieve flow balance without centralized control. The authors formalize the problem, specify a complete message-based coordination variant, and evaluate it through turn-based simulations across a wide parameter space to reveal how penetration level, human driver behavior, and variant choice influence balance and fairness. The study demonstrates that balanced flow can be achieved progressively as CAV penetration increases, and it provides a practical, upward-compatible framework that can bridge current capabilities to future generations, while acknowledging simplifications and outlining directions for dynamic bottlenecks and incentive mechanisms.

Abstract

Connected automated vehicles (CAV), which incorporate vehicle-to-vehicle (V2V) communication into their motion planning, are expected to provide a wide range of benefits for individual and overall traffic flow. A frequent constraint or required precondition is that compatible CAVs must already be available in traffic at high penetration rates. Achieving such penetration rates incrementally before providing ample benefits for users presents a chicken-and-egg problem that is common in connected driving development. Based on the example of a cooperative driving function for bottleneck traffic flows (e.g. at a roadblock), we illustrate how such an evolutionary, incremental introduction can be achieved under transparent assumptions and objectives. To this end, we analyze the challenge from the perspectives of automation technology, traffic flow, human factors and market, and present a principle that 1) accounts for individual requirements from each domain; 2) provides benefits for any penetration rate of compatible CAVs between 0 % and 100 % as well as upward-compatibility for expected future developments in traffic; 3) can strictly limit the negative effects of cooperation for any participant and 4) can be implemented with close-to-market technology. We discuss the technical implementation as well as the effect on traffic flow over a wide parameter spectrum for human and technical aspects.

Figures (7)

• Figure 1: Example situation at the bottleneck, with connected automated vehicles (CAVs) shown as shaded, and regular (human-driven) vehicles shown as outlines. Vehicles on the blocked lane ($b_1, ..., b_7$) have to wait for a vehicle on the free lane ($f_1, ..., f_8$) to yield the right of way. In this paper, an optimal decision-making algorithm for the CAVs is sought, while human drivers are assumed to act stochastically. The algorithm requires each CAV on the free lane to pick a distance threshold ${d}_{\textrm{\upshape{max}}}$, given here for the CAV $f_1$, specifying the maximum number of vehicles that the CAV would grant the right of way to.
• Figure 2: V2V messages exchanged at the example of Fig. \ref{['fig:example']}, with non-connected vehicles omitted. The front CAV on the free lane, $f_1$, broadcasts $m_{\textrm{\upshape{Invite}}}$, and receives the positions $m_{\textrm{\upshape{b}}}$ and $m_{\textrm{\upshape{f}}}$ from the blocked and free CAVs respectively. $f_1$ identifies $b_4$ as the furthest admissible CAV based on the set $\tilde{d}_{\textrm{\upshape{max}}} = 5$, and requests $b_4$ to return the right of way via $m_{\textrm{\upshape{RequestReturn}}}$. Once $b_4$ confirms via $m_{\textrm{\upshape{AcceptReturn}}}$, $f_1$ allows $b_2$ to drain without stopping via $m_{\textrm{\upshape{Clearance}}}$, and informs $b_7$ that it will not drain in this round via $m_{\textrm{\upshape{Dismiss}}}$. $f_1$ then stops and yields, likely flashing the high beam light to communicate the yielding to human drivers. Optionally (cf. Sec. \ref{['sec:counting']}) it counts draining vehicles (or measures elapsed time) until the opposite flow stops. Then $f_1$ drives into the bottleneck. In the non-counting variant, $f_3$ and $f_4$ will always receive permission to follow $f_1$ without yielding or sending $m_{\textrm{\upshape{Invite}}}$. In the counting variant (indicated by $*$), this will depend on the actions of the human drivers. If, for example, $b_3$ had decided to stop before the bottleneck, then only two vehicles from the blocked lane would have drained. In this case, $f_1$ would send $m_{\textrm{\upshape{Dismiss}}}$ to $f_3$ and $f_4$, advising them to offer $m_{\textrm{\upshape{Invite}}}$ before driving into the bottleneck. Either way, $f_8$ receives $m_{\textrm{\upshape{Dismiss}}}$.
• Figure 3: Example of the developed HMI based on ehrhardt2021gap, with compatible CAVs shown as solid cars, and other vehicles shown as outlines. The front CAV on the left yields the right of way to oncoming vehicles on the blocked lane. To communicate the situation to the human passengers inside via an HMI, the connection to the partner vehicle is indicated via signal bars, and a countdown circle indicates the maximum waiting time. The partner vehicle, the CAV on the right, has agreed to stop before entering the bottleneck, and thus return the right of way. Its passengers are also shown a connection to the respective partner vehicle, but an arrow indicates the "proceed until stop" status. A study on the impact of such an HMI design is given in ehrhardt2021gap.
• Figure 4: View of the scenario simulation in OCTANE showing a two-way street where one driving direction is blocked by a construction site. Vehicles in this direction have to drive on the opposite lane. In dense traffic, this requires an explicit stopping and yielding by the vehicles on the free lane.
• Figure 5: Extract of the simulation scenarios from a turn-based perspective of vehicle movements sliced by relevant actions ($\tau$) instead of time. Vehicles which have already passed the bottleneck are not shown. Numbered vehicles are compatible CAVs. As a measure of distance, $\tilde{d}$, the approximate number of cars to the bottleneck, is used and indicated. For vehicles on the free lane (left) their choice for $\tilde{d}_{\textrm{\upshape{max}}}$ is given. For vehicles on the blocked lane (right) the $\tilde{d}$ corresponding to their current position is indicated. Any unnumbered box is a vehicle that acts based on stochastic human behavior parameters. The scenario shows a compatible CAV yielding the right of way at $\tau = 3$ based on the acceptance of the blocked CAV at $\tilde{d} = 7$ to wait. This benefits the CAV in front of it ($\tilde{d} = 3$) which receives explicit permission to drive into the bottleneck. The rear vehicle waits as promised at $\tau = 10$ and returns the right of way to the free lane. It is granted the right of way by an obliging human driver at $\tau = 20$ before the next CAV on the free lane arrives at the front (which would also have yielded, based on its $\tilde{d}_{\textrm{\upshape{max}}} = 7$ and the availability of a blocked CAV at $\tilde{d}=6$.