In agile quadrotor flight, accurately modeling the varying aerodynamic drag forces encountered at different speeds is critical. These drag forces significantly impact the performance and maneuverability of the UAV, especially during high-speed maneuvers. Traditional control models based on first principles struggle to capture these dynamics due to the complexity and variability of aerodynamic effects, which are challenging to model accurately. To address these challenges, this study proposes a meta-learning-based control strategy for accurately modeling quadrotor dynamics under varying speeds, treating each velocity condition as an independent learning task with a specifically trained neural network to ensure precise dynamic predictions. The meta-learning framework rapidly generates task-specific parameters adapted to speed variations by solving an optimization problem and employs an online incremental learning strategy to integrate real-time data for continuous model updates, enhancing system robustness. Regularization is introduced to prevent overfitting and improve generalizability. The integration of the meta-learned model into Model Predictive Contouring Control (MPCC) allows the system to achieve optimal control across different velocity levels, ensuring efficient and accurate flight control even during sharp turns and high-speed maneuvers. Extensive simulations and real-world experiments confirm that the proposed algorithm maintains a high level of control precision despite the nonlinear effects of rapid speed changes, complex flight trajectories and wind disturbances. The results highlight the advantages of combining meta-learning with adaptive control strategies, providing a robust framework for quadrotors operating in diverse and dynamic environments.
The paper presents an air-assisted ground robotic autonomous exploration framework (AAGE), which leverages the high mobility and wide aerial perspective of Unmanned Aerial Vehicles (UAVs) to assist Unmanned Ground Vehicles (UGVs) in detailed exploration, enhancing exploration efficiency and improving the quality of point cloud collection in regions of interest (RoIs) in large-scale, unknown environments. In this framework, the UAV, equipped with an onboard RGB camera, rapidly surveys large unknown areas and generates a Bird's Eye View (BEV) to identify critical zones for UGV exploration. With prior information about the unexplored area's outline from the real-time shared BEV, the UGV can carry out more efficient and informed exploration from a global perspective. To maximize the utility of this prior information and optimize point cloud collection, a hierarchical exploration strategy and an attention mechanism are incorporated to guide the UGV's focus toward areas requiring detailed mapping, rather than broad, featureless regions. Real-world experiments validate the effectiveness of the framework, demonstrating significant improvements in exploration efficiency and point cloud collection compared to state-of-the-art methods. The results further show that even with a coarse BEV, the UGV's exploration efficiency is greatly enhanced.
Autonomous exploration of unknown environments is a challenging problem in robotic applications, especially in large-scale environments. As the size of the environment increases, the limited onboard resources of the robot hardly satisfy the memory overhead and computational requirements. As a result, it is challenging to respond quickly to the received sensor data, resulting in inefficient exploration planning. And it is difficult to comprehensively utilize the gathered environmental information for planning, leading to low-quality exploration paths. In this paper, a systematic framework tailored for unmanned aerial vehicles is proposed to autonomously explore large-scale unknown environments. To reduce memory consumption, a novel low-memory environmental representation is introduced that only maintains the information necessary for exploration. Moreover, a hierarchical exploration approach based on the proposed environmental representation is developed to allow for fast planning and efficient exploration. Extensive simulation tests demonstrate the superiority of the proposed method over current state-of-the-art methods in terms of memory consumption, computation time, and exploration efficiency. Furthermore, two real-world experiments conducted in different large-scale environments also validate the feasibility of our autonomous exploration system.
9 February 2026 in IEEE Robotics and Automation Letters
Accurate spatiotemporal calibration is crucial for effective data fusion in Visual-Inertial-LiDAR (VIL) systems. However, it remains challenging for setups involving omnidirectional cameras due to their ultra-wide Field-of-View (FoV) and severe distortion. Existing methods often rely on calibration targets or lack native support for such configurations. This letter presents OmniVIL-Calib, a target-free framework for the joint spatiotemporal calibration of an omnidirectional camera, IMU, and LiDAR. We propose an omnidirectional motion-flow model that generalizes the mapping-free paradigm to handle the non-linear geometric distortions inherent in ultra-wide FoV cameras and introduce a robust spherical cross-modal association to align visual and LiDAR edges. These constraints are jointly optimized to estimate all extrinsic parameters simultaneously. Validated on our self-collected OmniVIL dataset, the proposed approach achieves an average accuracy of 0.82 cm in translation and 0.13$^{\circ }$ in rotation. Experimental results show that our method achieves accuracy comparable to target-based benchmarks and exhibits superior performance compared to existing target-free solutions.
1 September 2026 in IEEE Transactions on Industrial Electronics
In dynamic and cluttered environments, quadrotors face significant challenges in simultaneously achieving precise trajectory tracking while maintaining robust safety guarantees due to time-varying disturbances and complex interactions with obstacles. Traditional control methods often fail to address these dual objectives simultaneously, with model-based approaches struggling to quantify uncertainty, and learning-based methods lacking formal safety assurances. To bridge this gap, this study proposes a learning-based safety-preserving quadratic control (LB-SPQC) framework that integrates adaptive learning with rigorous stability and safety constraints. A Gaussian process (GP)-based disturbance observer is developed, incorporating online hyperparameter optimization to enable real-time uncertainty quantification with probabilistic error bounds. The observer dynamically adapts to varying aerodynamic effects and environmental disturbances, ensuring accurate modeling across diverse flight conditions. In addition, a unified quadratic programming (QP) architecture is proposed, combining robust control Lyapunov functions (CLFs) for exponential stability and high-order control barrier functions (CBFs) for collision avoidance, while accounting for actuator saturation limits. This guarantees both tracking performance and safety. A computationally efficient sparse approximation scheme is implemented to reduce GP complexity through optimized inducing point selection, allowing real-time execution on embedded hardware. The framework is validated through extensive simulations and real-world flight experiments. Results demonstrate superior performance under challenging conditions, including wind disturbances and complex obstacle environments, showcasing LB-SPQC’s ability to provide a reliable and computationally efficient solution for quadrotor operations in dynamic scenarios.