Hierarchical Speed Planner for Automated Vehicles: A Framework for Lagrangian Variable Speed Limit in Mixed Autonomy Traffic

TL;DR

This work presents a hierarchical Speed Planner framework for Lagrangian variable speed limits in mixed autonomy traffic, combining a server-side planner with vehicle-side controllers to manage large-scale AV platoons. The system enhances traffic state estimation by predicting congestion fronts and fusing them with real-time probe-vehicle data, then designs per-vehicle target speeds via kernel smoothing, a learning-based buffer upstream of bottlenecks, and an optimization-based planner. A Transformer-based frontier predictor and a PDE-ODE traffic model underpin the control design, with an Actor-Critic RL policy guiding throughput optimization and an MPC benchmark provided for comparison. Field validation on MegaVanderTest with 100 AVs on I-24 MOTION demonstrates improvements in bottleneck throughput, fuel efficiency, and speed homogenization, while highlighting driver engagement and social acceptance as critical factors for real-world deployment.

Abstract

This paper introduces a novel control framework for Lagrangian variable speed limits in hybrid traffic flow environments utilizing automated vehicles (AVs). The framework was validated using a fleet of 100 connected automated vehicles as part of the largest coordinated open-road test designed to smooth traffic flow. The framework includes two main components: a high-level controller deployed on the server side, named Speed Planner, and low-level controllers called vehicle controllers deployed on the vehicle side. The Speed Planner designs and updates target speeds for the vehicle controllers based on real-time Traffic State Estimation (TSE) [1]. The Speed Planner comprises two modules: a TSE enhancement module and a target speed design module. The TSE enhancement module is designed to minimize the effects of inherent latency in the received traffic information and to improve the spatial and temporal resolution of the input traffic data. The target speed design module generates target speed profiles with the goal of improving traffic flow. The vehicle controllers are designed to track the target speed meanwhile responding to the surrounding situation. The numerical simulation indicates the performance of the proposed method: the bottleneck throughput has increased by 5.01%, and the speed standard deviation has been reduced by a significant 34.36%. We further showcase an operational study with a description of how the controller was implemented on a field-test with 100 AVs and its comprehensive effects on the traffic flow.

Figures (18)

• Figure 1: Hierarchical framework of the proposed VSL system: The Speed Planner module fetch inputs from the database to generate real-time target speed profile. Vehicle controllers get assigned targets speed via an API, together with local observation collected by on-board unit, as the input to decide the instant vehicle control command.
• Figure 2: Data pipeline for the Speed Planner: 1.For each update, past INRIX data from the database are fetched as the input of the prediction module. 2. Fetch the vehicle observations of the previous 1 minute. Fuse the INRIX prediction with vehicle observation to obtain the lane level TSE. 3. Smooth the obtained lane level TSE with the forward average kernel. 4. Input the smoothed TSE into the bottleneck identification module. 5. If there is any standing bottleneck identified, design the corresponding buffer segment in the smoothed TSE as the target speed profile. Else use the smoothed TSE as the target speed profile. 6. Publish.
• Figure 3: Prediction Module for TSE Enhancement: (a) INRIX heatmap on November 18th, 2022 (Left) and congestion frontier identification result (Right); (b) Network architecture of frontier predictor; (c) Speed filling to obtain TSE prediction.
• Figure 4: Training and field testing pipeline of the buffer design module.
• Figure 5: Obtain target speed profile: once a standing bottleneck is identified, the speed profile $v(t,x)$ will be converted to the density profile $\tilde{\rho}(t,x)$ by using a calibrated FD. RL policy selects the desirable density for the buffer area $\rho_b$ based on $\tilde{\rho}(t,x)$, which is the critical parameter to determine a desirable density profile $\rho^*(t,x)$. The target speed profile is obtained by converting $\rho^*(t,x)$ to the speed profile and taking the minimum value at each position.

Theorems & Definitions (5)

• remark 1
• definition 1: Weak solution
• theorem 1: Existence of solution
• proof : Sketch of the Proof for Theorem \ref{['th:existence']}
• definition 2: Riemann Solution