27 November 2024 in IEEE Transactions on Vehicular Technology
Cooperative decision-making of Connected Autonomous Vehicles (CAVs) presents a longstanding challenge due to its inherent nonlinearity, non-convexity, and discrete characteristics, compounded by the diverse road topologies encountered in real-world traffic scenarios. The majority of current methodologies are only applicable to a single and specific scenario, predicated on scenario-specific assumptions. Consequently, their application in real-world environments is restricted by the innumerable nature of traffic scenarios. In this study, we propose a unified optimization approach that exhibits the potential to address cooperative decision-making problems related to traffic scenarios with generic road topologies. This development is grounded in the premise that the topologies of various traffic scenarios can be universally represented as Directed Acyclic Graphs (DAGs). Particularly, the reference paths and time profiles for all involved CAVs are determined in a fully cooperative manner, taking into account factors such as velocities, accelerations, conflict resolutions, and overall traffic efficiency. The cooperative decision-making of CAVs is approximated as a mixed- integer linear programming (MILP) problem building on the DAGs of road topologies. This favorably facilitates the use of standard numerical solvers and the global optimality can be attained through the optimization. Case studies corresponding to different multi-lane traffic scenarios featuring diverse topologies are scheduled as the test itineraries, and the efficacy of our proposed methodology is corroborated.
16 July 2025 in IEEE Transactions on Vehicular Technology
Interactive decision-making and motion planning is essential for ensuring the safety and driving efficiency of autonomous vehicles. Game-based methods have demonstrated impressive performances in addressing these challenges due to their capability to explicitly model interactive behaviors. However, existing game-based approaches suffer from low computational efficiency owing to the numerical optimization required in continuous action spaces and the high dimensionality inherently present in multi-vehicle interaction scenarios. Also, accurately identifying the utility functions of opponent vehicles (OVs) from limited state observations remains a formidable challenge, which could result in conservative or even unsafe decisions by autonomous vehicles. To tackle these issues, a novel game-theoretic decision-making and motion planning framework is proposed. Specifically, the multi-vehicle interaction is transformed to a pairwise leader-follower game and further modelled by trajectory trees constructed with discrete action sets. To accurately compute the action values of these trajectory trees, we identify the utility functions of OVs from state observations by solving the elaborated restored optimal control problem. The optimal strategy for the trajectory trees is derived using dynamic programming framed within the context of a sequential game, which ensures that the varying behaviors of OVs are effectively considered. The proposed method is comprehensively validated in two challenging interactive scenarios, and the results show that the trajectory tree-based interaction modeling approach significantly enhances the computational efficiency of game-based methods. Additionally, the identified utility functions of OVs provide a better interpretation of their observed behaviors and ensure safer decision-making and motion planning for autonomous vehicles.
2 December 2024 in IEEE Transactions on Vehicular Technology
This paper introduces a local planner that synergizes the decision making and trajectory planning modules towards autonomous driving. The decision making and trajectory planning tasks are jointly formulated as a nonlinear programming problem with an integrated objective function. However, integrating the discrete decision variables into the continuous trajectory optimization leads to a mixed-integer programming (MIP) problem with inherent nonlinearity and nonconvexity. To address the challenge in solving the problem, the original problem is decomposed into two sub-stages, and a two-stage optimization (TSO) based approach is presented to ensure the coherence in outcomes for the two stages. The optimization problem in the first stage determines the optimal decision sequence that acts as an informed initialization. With the outputs from the first stage, the second stage necessitates the use of a high-fidelity vehicle model and strict enforcement of the collision avoidance constraints as part of the trajectory planning problem. We evaluate the effectiveness of our proposed planner across diverse multi-lane scenarios. The results demonstrate that the proposed planner simultaneously generates a sequence of optimal decisions and the corresponding trajectory that significantly improves driving performance in terms of driving safety and traveling efficiency as compared to alternative methods. Additionally, we implement the closed-loop simulation in CARLA, and the results showcase the effectiveness of the proposed planner to adapt to changing driving situations with high computational efficiency.
14 November 2025 in IEEE Transactions on Vehicular Technology
Integrating General Models (GMs) such as Large Language Models (LLMs), with Specialized Models (SMs) in autonomous driving tasks presents a promising approach to mitigating challenges in data diversity and model capacity of existing specialized driving models. However, this integration leads to problems of asynchronous systems, which arise from the distinct characteristics inherent in GMs and SMs. To tackle this challenge, we propose NetRoller, an adapter that incorporates a set of novel mechanisms to facilitate the seamless integration of GMs and specialized driving models. Specifically, our mechanisms for interfacing the asynchronous GMs and SMs are organized into three key stages. NetRoller first harvests semantically rich and computationally efficient representations from the reasoning processes of LLMs using an early stopping mechanism, which preserves critical insights on driving context while maintaining low overhead. It then applies learnable query embeddings, nonsensical embeddings, and positional layer embeddings to facilitate robust and efficient cross-modality translation. At last, it employs computationally efficient Query Shift and Feature Shift mechanisms to enhance the performance of SMs through few-epoch fine-tuning. Based on the mechanisms formalized in these three stages, NetRoller enables specialized driving models to operate at their native frequencies while maintaining situational awareness of the GM. Experiments conducted on the nuScenes dataset demonstrate that integrating GM through NetRoller significantly improves human similarity and safety in planning tasks, and it also achieves noticeable precision improvements in detection and mapping tasks for end-to-end autonomous driving. The code and models are available at https://github.com/Rex-sys-hk/NetRoller.
24 August 2026 in Unmanned Systems