Planning safe and smooth trajectories for multiple autonomous ground vehicles (MAGVs) in a complex dynamic unstructured environment is a fundamental and challenging task. In this paper, a novel unified framework integrating trajectory planning and motion optimization (TPMO) is proposed based on spatio-temporal safety corridor (STSC), which guarantees collision avoidance and trajectory smoothness. The proposed TPMO framework consists of two parts. The first part is to establish the STSC for each AGV based on the mixed integer quadratic programming (MIQP) algorithm. The proposed STSC method ensures collision avoidance in the environment of static and dynamic obstacles, and provides a longitudinal and lateral coupled trajectory (LLCT) for trajectory planning. The second part is to design a motion optimization methodology, which considers the constraints of AGV geometry as well as longitudinal and lateral coupled motion characteristics. Moreover, our formulation provides a theoretical guarantee that the entire trajectory is optimal under collision avoidance. Finally, the proposed TPMO framework is applied to solve the optimal cooperative trajectory and motion planning problem of MAGVs in a near-natural simulation and real vehicle environments, validating the proposed framework's effectiveness and practicality.

This letter studies the trajectory tracking and motion control problems of unmanned ground vehicles (UGVs). A model predictive control and robust control (MPC-RC) framework for UGVs is proposed to improve tracking accuracy, yaw stability and robustness in a modular fashion without introducing complexity into controller. The trajectory tracking problem and three-dimensional phase trajectory planning with high stability of a vehicle motion can be performed in the model predictive control design simultaneously. Also, combining the advantages of linear matrix inequality, sliding mode control, and back-stepping control law, three robust motion controllers can track the generated three-dimensional phase trajectory steadily so that the UGV motion stability is guaranteed. The robust performance is guaranteed through considering model uncertainties and terra-aerodynamic disturbances in robust controllers. Sufficient conditions for closed-loop stability under the diverse robust factors are provided by the Lyapunov method analytically, which ensures the series system's feasibility. The results of simulations on MATLAB-Carsim platform demonstrate that the proposed controller can significantly enhance tracking accuracy, motion stability, and robustness compared to the existing methods, which guarantees the feasibility and capability of driving in a nonlinear extreme scenario.

Generating safe, smooth, and efficient trajectories is a fundamental and difficult task for multiple unmanned ground vehicles (MUGVs) in unstructured narrow interaction scenarios (UNIS). In this paper, a coordinated behavior planning and trajectory planning (BPTP) framework is proposed to deal with the MUGVs collision-prone problem in UNIS, which consists of interactive behavior planning layer and an optimization-based trajectory planning layer. The behavior planning layer is designed based on the mixed integer quadratic programming (MIQP) and breadth-first search (BFS) approach to generate collision-free trajectory clusters in the spatio-temporal state. The selected coarse trajectory fully considers the continuity of the trajectory, the non-collision and the driving efficiency between MUGVs. For trajectory planning, a spatio-temporal motion corridor (STMC) approach is proposed based on MIQP and optimization approach to model the constraints posed by complex unstructured environments in a unified way. Based on STMC, a safe and smooth trajectory is optimized, conforming to the decision provided by the behavior planner. Finally, the proposed BPTP framework is validated in both simulations and real-vehicle, and experimental results demonstrate that the proposed framework achieves human-like driving behavior in UNIS safely and smoothly compared to the existing planning methods.

High-efficiency, real-time and precise formation trajectory tracking control (FTTC) is a fundamental and difficult task for unmanned tracked vehicles (UTVs) in unstructured variable curvature scenarios. It is quite challenging and time-consuming for FTTC optimization on UTVs, due to its strongly nonlinear constraint, poor real-time and computation complexity. In this paper, a coupling multi-objective iterative distributed model predictive control (CMOI-DMPC) approach is proposed to address the above difficulties. The proposed CMOI-DMPC approach consists of two steps. The first step is to build a linearized cross-coupled instantaneous centers of rotation (ICR) kinematics model of the UTV to address the strongly nonlinear constraints. The second step is to design a multi-objective iterative optimization DMPC strategy to solve the problems of time-consuming computation and low tracking accuracy. Moreover, the sufficient condition on ensuring closed-loop stability is demonstrated by Lyapunov theorem. The proposed CMOI-DMPC approach is also applied to solve an optimal FTTC problem of UTVs in near-natural simulation environment and real vehicle environment, which validates the effectiveness and practicality of the proposed CMOI-DMPC approach. A near-natural simulation video of our approach was released here: https://space.bilibili.com/430786265 .

Large-scale 3D scene reconstruction and novel view synthesis are vital for autonomous vehicles, especially utilizing temporally sparse LiDAR frames. However, conventional explicit representations remain a significant bottleneck towards representing the reconstructed and synthetic scenes at unlimited resolution. Although the recently developed neural radiance fields (NeRF) have shown compelling results in implicit representations, the problem of large-scale 3D scene reconstruction and novel view synthesis using sparse LiDAR frames remains unexplored. To bridge this gap, we propose a 3D scene reconstruction and novel view synthesis framework called parent-child neural radiance field (PC-NeRF). Based on its two modules, parent NeRF and child NeRF, the framework implements hierarchical spatial partitioning and multi-level scene representation, including scene, segment, and point levels. The multi-level scene representation enhances the efficient utilization of sparse LiDAR point cloud data and enables the rapid acquisition of an approximate volumetric scene representation. With extensive experiments, PC-NeRF is proven to achieve high-precision novel LiDAR view synthesis and 3D reconstruction in large-scale scenes. Moreover, PC-NeRF can effectively handle situations with sparse LiDAR frames and demonstrate high deployment efficiency with limited training epochs. Our approach implementation and the pre-trained models are available at https://github.com/biter0088/pc-nerf .