The guidance problem of a confrontation between an interceptor, a hypersonic vehicle, and an active defender is investigated in this paper. As a hypersonic multiplayer pursuit-evasion game, the optimal guidance scheme for each adversary in the engagement is proposed on the basis of linear-quadratic differential game strategy. In this setting, the angle of attack is designed as the output of guidance laws, in order to match up with the nonlinear dynamics of adversaries. Analytical expressions of the guidance laws are obtained by solving the Riccati differential equation derived by the closed-loop system. Furthermore, the satisfaction of the saddle-point condition of the proposed guidance laws is proven mathematically according to the minimax principle. Finally, nonlinear numerical examples based on 3-DOF dynamics of hypersonic vehicles are presented, to validate the analytical analysis in this study. By comparing different guidance schemes, the effectiveness of the proposed guidance strategies is demonstrated. Players in the engagement could improve their performance in confrontation by employing the proposed optimal guidance approaches with appropriate weight parameters.
The cooperative active defense guidance problem for a spacecraft with active defense is investigated in this paper. An engagement between a spacecraft, an active defense vehicle, and an interceptor is considered, where the target spacecraft with active defense will attempt to evade the interceptor. Prior knowledge uncertainty and observation noise are taken into account simultaneously, which are vital for traditional guidance strategies such as the differential-game-based guidance method. In this set, we propose an intelligent cooperative active defense (ICAAI) guidance strategy based on deep reinforcement learning. ICAAI effectively coordinates defender and target maneuvers to achieve successful evasion with less prior knowledge and observational noise. Furthermore, we introduce an efficient and stable convergence (ESC) training approach employing reward shaping and curriculum learning to tackle the sparse reward problem in ICAAI training. Numerical experiments are included to demonstrate ICAAI’s real-time performance, convergence, adaptiveness, and robustness through the learning process and Monte Carlo simulations. The learning process showcases improved convergence efficiency with ESC, while simulation results illustrate ICAAI’s enhanced robustness and adaptiveness compared to optimal guidance laws.
17 October 2025 in IEEE Robotics and Automation Letters
Autonomous exploration in large-scale, unknown environments remains a significant challenge in mobile robotics. In this letter, we propose a scalable exploration framework that integrates heterogeneous topological representations, lightweight global-local graph reasoning, and reinforcement learning. Our framework is computationally efficient, comprehensively considers global spatial context, and generalizes effectively across diverse environmental scenarios. We introduce a compact topological abstraction to encode crucial spatial and semantic information, substantially reducing map complexity. A novel hybrid inference module, combining linear global attention with local graph convolutions, effectively integrates long-range exploration with comprehensive local coverage. We further design a minimalist reward function paired with a curriculum learning to ensure stable training and enhanced generalization. A viewpoint-based action masking mechanism further refines the action space, accelerating learning convergence. Extensive simulations demonstrate our method consistently surpasses state-of-the-art baselines, achieving up to 14.5% shorter exploration time and 18.8% reduced path length while maintaining low computational overhead. Real-world experiments further validate the practical effectiveness of our approach for robotic exploration tasks.
9 April 2026 in IEEE Robotics and Automation Letters
Reinforcement learning (RL) has renewed interest in end-to-end robot navigation, yet dynamic, crowd-like scenes remain difficult due to partial observability and the fragility of online tracking. We introduce GPTS-Nav, a Graph-Privileged Teacher-Student RL framework for LiDAR-only navigation. A graph-aware teacher is trained using single-step privileged motion graphs that explicitly include ground-truth velocities, forming a complete Markov state. In contrast, a sensor-only student relies purely on temporal observation history to implicitly infer unobservable dynamics. To enhance distillation fidelity and physical execution, a Gated LoRA Fusion module aligns student temporal features with teacher graph privileges, while an Action Chunking Transformer (ACT) predicts action chunks to filter jitter and surpass teacher trajectory smoothness. In large-scale simulations, GPTS-Nav consistently surpasses strong optimization-based and prior RL-based methods in success rate and efficiency. Furthermore, quantitative real-world experiments on a wheeled–biped platform validate its robust performance across narrow passages and uncooperative pedestrian zones, demonstrating safe, smooth navigation and readiness.
1 July 2026 in IEEE Robotics and Automation Letters