Lateral lane-changing and longitudinal car-following behavior decision-making is crucial for the implementation of autonomous driving (AD) in complex and dynamic traffic environment. Reinforcement learning (RL) has been increasingly applied to these behaviors due to its powerful environmental adaptability. However, typical RL-based research focused solely on one directional decision-making or output discrete actions in both lateral and longitudinal directions, which is not sufficiently flexible for AD. In view of the close interaction between lane-changing and car-following tasks, this study proposed an integrated framework to realize coupled decision-making using RL. Specifically, double deep Q-network (DDQN) was used to make the lateral lane change command decision-making, and twin delay deep deterministic policy gradient (TD3) used to make longitudinal car-following acceleration decision-making. The integration of DDQN and TD3 was deployed to tackle the problem of discrete-continuous hybrid action space in RL. The lateral decision-making was integrated into the actor-critic network for longitudinal decision-making to generate accurate acceleration in lane-changing or lane-keeping conditions. A series of simulation experiments were conducted in typical two-lane straight driving scenarios, validating the feasibility and applicability of the proposed integrated framework. The results reveal that the RL-based coupled decision-making helps enhance the implementation of AD.

Applications of autonomous mining trucks (AMTs) at open-pit mining sites are becoming a trend. Improving the driving efficiency of AMTs in dumping area enables them to quickly engage in the next loading task, thereby increasing mining productivity. In practice, this specific scenario presents typical challenges: 1) numerous path intersections and overlaps lead to severe AMT conflicts; 2) spatiotemporal conflicts are difficult to resolve and easy to cause deadlocks with increasing dumping demand; 3) the motion planning model is difficult to obtain solutions in real time. To resolve these problems, this study proposes a multi-AMT cooperative motion planning framework for dumping area, where the mine's cloud platform centrally computes terminal conditions for the speed planning of each AMT to coordinate conflicts, and each AMT performs speed planning in a distributed manner to ensure driving safety. In terms of centralized computing, a mixed-integer linear program (MILP) model is developed, and a grouped Monte Carlo tree search (GMCTS) model, inspired by connected component analysis in graph theory, is proposed to solve the MILP model within limited iterations. Simulation experiments of real dumping areas show that the proposed method allows to avoid deadlocks and enhance AMTs' operational efficiency within desirable computation time.

Connected and autonomous vehicles (CAVs) have been deployed rapidly at unstructured intersections. To prevent frequent stops and ensure safety, CAVs should pass through the intersection within recommended departure time windows while adhering to specific motion constraints on terrains with irregular boundaries and slopes. It leads to a multi-vehicle trajectory planning (MVTP) problem at intersection areas, for which a mixed-integer nonlinear programming (MINLP) model with non-differentiable multiple if-else expressions is applicable but difficult to solve efficiently by existing optimizers. To this end, this study proposes a centralized-distributed computing framework to tackle this problem in an efficient manner. Specifically, at the centralized level, an improved conflict-based search (CBS) and hybrid A* algorithm is used to generate coarse trajectories satisfying departure time windows constraints whereas handling unknown if-else expressions. At the distributed level, each CAV performs a simple NLP-based trajectory optimization, with collision-avoidance ensured through linear corridor constraints. Experimental validation and comparisons with benchmarks demonstrate that the effectiveness and efficiency of the proposed framework.