Energy-efficient predictive control for connected, automated driving under localization uncertainty

TL;DR

The paper tackles energy-efficient longitudinal control for connected automated vehicles under localization uncertainty on urban routes with traffic lights and a front vehicle. It proposes a unified, data-driven MPC that learns a terminal cost $V_f(\cdot)$ and a terminal set $\mathcal{X}_f$ from closed-loop data, while using a short-horizon MPC with a stochastic-terminal-cost approximation to handle uncertainty. The approach replaces hierarchical planning with a single controller, achieves up to 19–19.4% energy savings versus strong baselines, and demonstrates robust performance in simulation and vehicle-in-the-loop experiments. The method's practical impact lies in enabling energy-efficient CAV operation under imperfect positioning, with potential extensions to dynamic traffic and multi-vehicle coordination via V2V.

Abstract

This paper presents a data-driven Model Predictive Control (MPC) for energy-efficient urban road driving for connected, automated vehicles. The proposed MPC aims to minimize total energy consumption by controlling the vehicle's longitudinal motion on roads with traffic lights and front vehicles. Its terminal cost function and terminal constraints are learned from data, which consists of the closed-loop state and input trajectories. The terminal cost function represents the remaining energy-to-spend starting from a given terminal state. The terminal constraints are designed to ensure that the controlled vehicle timely crosses the upcoming traffic light, adheres to traffic laws, and accounts for the front vehicles. We validate the effectiveness of our method through both simulations and vehicle-in-the-loop experiments, demonstrating 19% improvement in average energy efficiency compared to conventional approaches that involve solving a long-horizon optimal control problem for speed planning and employing a separate controller for speed tracking.

Figures (7)

• Figure 1: The black line is the bounded support $2LN\mathcal{W}$. All dots are the discretized noises. The blue dots denote the recorded noises while the red dots denote vertices of the bounded support $2LN\mathcal{W}$.
• Figure 2: Defintion of $t_\text{red}$ and $t_\text{green}$
• Figure 3: Illustration of the terminal constraints $\mathcal{S}_{t_\text{red}}$ and $\mathcal{P}_{t_\text{green}}$: the constraint $\hat{\mathbf{x}}_{N|k} \in \mathcal{S}_{t_\text{red}}$ is to ensure that the vehicle stays behind the traffic light until $t_\text{red}+N$ steps, while the constraint $\hat{\mathbf{x}}_{N|k} \in \mathcal{P}_{t_\text{green}}$ is to ensure that the vehicle passes the traffic light within $t_\text{green}+N$ steps.
• Figure 4: Total Energy Consumption Decreases with Increasing Data Size. $100$ Monte Carlo simulations for each task iteration.
• Figure 5: Simulation results when $v_k^{pv}=7.5$m$/$s. (a) and (b): Vehicle speed histories: Comparison between Proposed Algorithm and Algorithm in bae2019VIL. Closed-loop states and Predicted or Planned states at four different time steps. (c) Distance to front Vehicle: Comparison between two algorithms. (d) Energy Consumption: Comparison between two algorithms.

Theorems & Definitions (11)

• Proposition 1
• proof
• Remark 1
• Remark 2
• Remark 3
• Remark 4
• Theorem 1
• proof
• Theorem 2
• proof