TL;DR
This paper introduces a novel bi-level control system (RCS) for Connected and Automated Vehicles (CAVs) at roundabouts. By coordinating vehicle merging through higher-level clustering and lower-level optimization, the RCS significantly improves traffic flow, reduces travel time and fuel consumption compared to traditional systems.
What Problem Does This Paper Address?
Traditional roundabouts, while generally safer than signalized intersections, can still lead to significant traffic congestion, delays, and increased fuel consumption, especially with increasing vehicle volumes. This is primarily due to the uncoordinated nature of human driving, leading to stop-and-go traffic, inefficient merging, and suboptimal use of roundabout capacity. The paper aims to mitigate these issues by leveraging the capabilities of Connected and Automated Vehicles (CAVs) to achieve smooth, safe, and efficient merging at roundabouts.
Methods
The proposed solution is a Bi-Level Coordinated Merging System (RCS) managed by a central Roundabout Coordination Unit (RCU).
- Higher Level Coordination (Clustering Zone): This level focuses on proactively managing traffic flow before vehicles reach the immediate merging point. The RCU uses traffic flow information (entry and circulating rates) to determine if vehicles in the "clustering zone" (60-200m from the roundabout) should form vehicle clusters (platoons). This clustering aims to create more organized arrival patterns, reducing the need for individual vehicles to make aggressive maneuvers or stop. The RCU calculates recommended speeds for vehicles to join or form these clusters.
- Lower Level Coordination (Merging Execution Zone): Once vehicles are in the "merging execution zone" (0-60m from the roundabout), this level handles the precise merging. It employs a Receding Horizon Control (RHC) approach to solve a combinatorial optimization problem. This optimization determines the optimal merging sequence and timing for vehicles from both entry and circulating lanes to minimize travel time while ensuring safety. The RCU then communicates these optimal merging times to the vehicles, which use their local controllers (e.g., Adaptive Cruise Control) to adjust their acceleration accordingly.
The paper also mentions the use of the MOBIL (Minimizing Overall Braking Induced by Lane Change) model for a final safety check before merging.
Datasets
Not discussed in the paper. The evaluation was performed using microscopic simulations in MATLAB.
Key Insights and Ideas
- Clustering for Efficiency: The idea of forming vehicle clusters before the merging point is a key innovation. Instead of just reacting at the merge, proactive grouping of vehicles smooths out arrivals and reduces the complexity of individual merging decisions.
- Bi-Level Optimization: Separating the problem into higher-level (clustering) and lower-level (sequential optimization) coordination allows for a more manageable and real-time implementable solution. The higher level simplifies the lower level's task.
- Receding Horizon Control for Real-Time: Using RHC for the combinatorial optimization problem is crucial for real-time application. It allows the system to continuously re-optimize based on the current state, making it robust to dynamic traffic conditions.
- Roundabouts as a Unique Challenge: While many ITS solutions focus on signalized intersections, roundabouts present a distinct challenge due to their continuous flow and yielding requirements, making coordinated merging particularly impactful.
Key Results
The proposed RCS demonstrated significant improvements over the Traditional Roundabout System (TRS) in simulations:
- Case 1 (Balanced Traffic Flow):
- Average travel time reduced by up to 44.28%.
- Average idling time reduced by up to 36.20%.
- Average velocity increased by up to 11.20%.
- Average fuel consumption reduced by up to 11.20%.
- Case 2 (Unbalanced Traffic Flow):
- Average travel time reduced by up to 45.90%.
- Average idling time reduced by up to 38.34%.
- Average velocity increased by up to 11.92%.
- Average fuel consumption reduced by up to 11.92%.
These improvements were particularly noticeable at higher traffic flow rates, approaching capacity.
Limitations and Future Work
- 100% CAV Penetration: The current study assumes a 100% penetration rate of CAVs. The authors acknowledge that investigating mixed traffic environments (CAVs and traditional human-driven vehicles) is a crucial next step, though challenging due to the difficulty in modeling human behavior.
- Communication Delay: While the study assumes reliable communication, the authors note that communication delays can impact CAV performance. They suggest that their current gap settings are larger than those for CACC systems, making them less susceptible, but acknowledge that further study on delay magnitude is needed.
- Single-Lane Roundabouts: The current simulation is for a single-lane roundabout. Future work will extend the model to multi-lane roundabouts.
- Distributed MPC: The scheme could potentially be extended using distributed Model Predictive Control (MPC) for individual vehicles.
- Generalization to Intersections: The proposed bi-level approach for coordination could be adapted for traditional intersections.
Summary updated