Gold nanoparticles, serving as a photothermal agent, are often embedded in tumor tissue to enhance the efficacy of laser hyperthermia. This paper investigates the thermomechanical response of a two-layer semi-infinite biological tissue composed of tumor tissue embedded with gold nanoparticles and normal tissue under laser irradiation. By incorporating the Pennes bioheat equation, a fractional three-phase-lag (FTPL) thermoelastic model is formulated to address the problem. The gold nanoparticles affect the laser absorption and scattering at the tumor site. The exact analytical solution of the fractional thermomechanical coupling problem is obtained by coupling the state-space method with the Laplace transformation. The influence of the FTPL model parameters, the convective cooling coefficient at the boundary, the gold nanoparticle concentration, and the laser parameters on the thermomechanical behavior are analyzed. This study establishes a unified thermomechanical coupling solution framework via the state-space method for laser hyperthermia, and provide valuable insights for the application and development of gold nanoparticles in biological thermal therapy.
Achieving stable and accurate estimation of failure probability with minimal evaluations of computationally expensive numerical models remains a fundamental challenge in reliability analysis. To overcome the limitations of existing surrogate-based and radial sampling approaches that rely on a prespecified initial reliability index and may exhibit instability in failure boundary learning, this paper proposes an adaptive (1-shell active learning Kriging method referred to as AK-AS. The proposed method employs an adaptive (1-shell strategy in which the reliability index is progressively identified and updated using Kriging model predictions, thereby eliminating the need for a user-defined initial reliability index and reducing the candidate sample space. Within this framework, two learning functions with complementary roles are introduced for adaptive surrogate refinement. The composite boundary-guided sampling learning function (CBS) enables lightweight correction of the initial Kriging model to support stable identification of the initial reliability index, while the radial penalty sampling learning function (RPS) focuses on refinement within narrow regions between adjacent (1-shells to suppress numerical oscillations and improve accuracy near the failure boundary. Numerical examples demonstrate that the proposed method achieves enhanced computational efficiency and estimation stability for problems involving small failure probabilities, multiple failure modes, and highly nonlinear limit-state functions.
4D millimeter-wave radar provides high-precision ranging capability and exhibits strong robustness under adverse weather and low-visibility conditions, but its point clouds are relatively sparse and suffer from severe elevation-angle measurement noise. Monocular cameras, by contrast, provide rich semantic information and high recall, yet are fundamentally limited by scale ambiguity. To exploit the complementary characteristics of these two sensors, this paper proposes a radar-camera fusion 3D multi-object tracking framework that does not rely on complex 3D annotated data. First, on the radar signal-processing side, a Gaussian distribution-based adaptive angle compression method and IMU-based velocity compensation are introduced to effectively suppress measurement noise, and an improved DBSCAN clustering scheme with recursive cluster splitting and historical static-box guidance is employed to generate high-quality radar detections. Second, a disparity-domain metric depth recovery method is proposed. This method uses filtered radar points as sparse metric anchors, performs robust fitting with RANSAC, and applies Kalman filtering for temporal smoothing, thereby converting the relative depth output of the visual foundation model Depth Anything V2 into metric depth. Finally, a hierarchical fusion strategy is designed at both the detection and tracking levels to achieve stable cross-modal state association. Experimental results on a self-collected dataset show that the proposed method achieves an overall MOTA of 77.93%, outperforming single-modality baselines and other comparison methods by 11 to 31 percentage points. This study provides an effective solution for low-cost and robust environment perception in complex dynamic scenarios.
Symbolic regression is an explainable machine learning approach to data analysis, which aims to distill ana lytical data-driven models that are represented as mathematical expressions by optimizing both the parameters and structure. Commonly used genetic programming based symbolic regression algorithms often suffer from low efficiency and the effects of blind search, model bloat and diversity loss when dealing with massive data. This pa per improves the performance of genetic programming based symbolic regression algorithms via handling these effects, task parallelism and hardware acceleration. We propose an asynchronous parallel framework based on multiobjective genetic programming algorithms, in which problem decomposition and elite sharing strategies are used to quickly produce various high-quality models and crosswise evolve elite models among different subtasks. In this way, such cross-evolution of elite models among subtasks can improve the model generalization, evolv ing efficiency and population diversity. Moreover, a tailor-made hybrid CPU-GPU acceleration strategy based on the compute unified device architecture is designed for the proposed algorithm by offloading model evalua tions and, reducing the memory usage pressure and the communications between the CPU and GPUs. Ablation and comparison experiments are conducted to evaluate the proposed framework. The results demonstrate the positivity and effectiveness in data depiction, bloat-control and time-efficiency of the proposed framework. The proposed algorithm framework achieves 6.5-12.4 & times; speedup for ground-truth problems and 5.1-13.4 & times; for black-box problems with CPU implementations respectively and, accordingly 6.1-365.5 & times; and 4.4-188.4 & times; speedup with CPU-GPU implementations against the standard multiobjective genetic programming based symbolic regression algorithm in terms of average model exploration rate. Accelerated by two GPUs, it achieves 2.6-28.2 & times; and 2.6-15.2 & times; speedup for the proposed asynchronous parallel approach on the ground-truth and black-box benchmarks with large datasets respectively.
Cooperative multi-UAV search requires jointly optimizing wide-area coverage, rapid target discovery, and endurance under sensing and motion constraints. Resolving this coupling enables scalable coordination with high data efficiency and mission reliability. We formulate this problem as a discounted Markov decision process on an occupancy grid with a cellwise Bayesian belief update, yielding a Markov state that couples agent poses with a probabilistic target field. On this belief-MDP we introduce a segment-conditioned latent-intent framework, in which a discrete intent head selects a latent skill every K steps and an intra-segment GRU policy generates per-step control conditioned on the fixed intent; both components are trained end-to-end with proximal updates under a centralized critic. On the 50 & times; 50 grid, coverage and discovery convergence times are reduced by up to 48% and 40% relative to a flat actor-critic benchmark, and the aggregated convergence metric improves by about 12% compared with a stateof-the-art hierarchical method. Qualitative analyses further reveal stable spatial sectorization, low path overlap, and fuel-aware patrolling, indicating that segment-conditioned latent intents provide an effective and scalable mechanism for coordinated multi-UAV search.