Yiguang Hong

Rapid cooperative detection of stationary targets by multiple unmanned aerial vehicles (UAVs) is important in time-critical missions such as search and rescue. However, the online coordination of probabilistic inference, distributed communication, and detection–motion decisions under local information remain challenging when targets exhibit spatial correlations that existing methods typically neglect. To address this challenge, we develop a distributed event-driven Bayesian search framework for stationary, spatially correlated targets at unknown locations. The framework couples three components. A pairwise spatial model and a distance-dependent Neyman–Pearson detector yield a Bayesian belief update whose unclipped product form is order-invariant to event-processing sequence. A distributed selective flooding algorithm propagates only positive detection events, achieving finite-time event-set consensus over connected graphs while avoiding full-map exchange. A decoupled detection–motion planner exhausts high-belief cells within each UAV’s field of view before selecting a waypoint that balances surrogate detection probability against travel cost, with responsibility regions dynamically renegotiated among neighbors when local high-value cells are depleted. In numerical experiments, the proposed method achieved zero uncoordinated repeat detection in all simulations and significantly reduced first-discovery coverage relative to static-partition and no-communication baselines, while adapted external baselines required 90-fold and 6-fold larger communication payloads and had nonzero repeat-detection rates. The framework thus occupies a specific tradeoff point of zero revisit, sparse communication, and early discovery gain in scenes where targets span multiple UAV search regions.

Reinforcement Learning (RL) has demonstrated excellent decision-making potential in platoon coordination problems. However, due to the variability of coordination goals, the complexity of the decision problem, and the time-consumption of trial-and-error in manual design, finding a well performance reward function to guide RL training to solve complex platoon coordination problems remains challenging. In this paper, we formally define the Platoon Coordination Reward Design Problem (PCRDP), extending the RL-based cooperative platoon coordination problem to incorporate automated reward function generation. To address PCRDP, we propose a Large Language Model (LLM)-based Platoon coordination Reward Design (PCRD) framework, which systematically automates reward function discovery through LLM-driven initialization and iterative optimization. In this method, LLM first initializes reward functions based on environment code and task requirements with an Analysis and Initial Reward (AIR) module, and then iteratively optimizes them based on training feedback with an evolutionary module. The AIR module guides LLM to deepen their understanding of code and tasks through a chain of thought, effectively mitigating hallucination risks in code generation. The evolutionary module fine-tunes and reconstructs the reward function, achieving a balance between exploration diversity and convergence stability for training. To validate our approach, we establish six challenging coordination scenarios with varying complexity levels within the Yangtze River Delta transportation network simulation. Comparative experimental results demonstrate that RL agents utilizing PCRD-generated reward functions consistently outperform human-engineered reward functions, achieving an average of 10% higher performance metrics in all scenarios.

Developing world models that understand complex physical interactions is essential for advancing robotic planning and simulation.However, existing methods often struggle to accurately model the environment under conditions of data scarcity and complex contact-rich dynamic motion.To address these challenges, we propose ContactGaussian-WM, a differentiable physics-grounded rigid-body world model capable of learning intricate physical laws directly from sparse and contact-rich video sequences.Our framework consists of two core components: (1) a unified Gaussian representation for both visual appearance and collision geometry, and (2) an end-to-end differentiable learning framework that differentiates through a closed-form physics engine to infer physical properties from sparse visual observations.Extensive simulations and real-world evaluations demonstrate that ContactGaussian-WM outperforms state-of-the-art methods in learning complex scenarios, exhibiting robust generalization capabilities.Furthermore, we showcase the practical utility of our framework in downstream applications, including data synthesis and real-time MPC.

In this paper, we consider how a selfish agent manipulates the optimal value of multi-agent optimization by local false information. First, we formulate a distributed optimization problem for multi-agent systems with a selfish agent that aims to shift the global optimal state to its preference via local information manipulation. Then, based on a well-known distributed proportion-integral optimization, the deception strategies are proposed, and the exponential convergence of the manipulated algorithms is proved. Moreover, for any given deviation tolerated by other agents, the selection of the deception parameter is provided as well.

This paper addresses the nonovershooting control problem for strict-feedback nonlinear systems with unknown control direction. We propose a method that integrates extremum seeking with Lie bracket-based design to achieve approximately nonovershooting tracking. The approach ensures that arbitrary reference trajectories can be tracked from below for any initial condition, with the overshoot reducible to arbitrarily small levels through parameter tuning. The method further provides a mechanism for enforcing high-relative-degree nonovershooting constraints in safety-critical scenarios involving unknown control directions.