纽约大学

Decentralized control has been a central focus in the management of large-scale urban traffic networks. Among various strategies, the Max-Pressure algorithm has emerged as a leading method for decentralized signal control. Notably, it operates without requiring prior knowledge of traffic demand, while implicitly assuming that arrival processes are admissible within the network’s stability region. Theoretical guarantees, including throughput optimality, are established under idealized assumptions that neglect signal control constraints and are largely inherited from its origins in communication networks. This study aims to characterize network capacity by defining the stability region for each intersection under full implementation of the Max-Pressure policy, while accounting for realistic traffic signal constraints such as inter-green periods and minimum green activation times. We focus on simple two-phase intersections, for which the signal stage sequence is fixed by design and does not require explicit sequencing constraints. As a foundational step, we first analyze the Max-Pressure controller in an isolated intersection setting. Using both analytical derivations and simulation, we evaluate the impact of temporal constraints - specifically inter-green times and minimum green durations - on the controller’s ability to maintain system stability. Our results demonstrate that once these practical constraints are introduced, the classical throughput-optimality guarantees no longer hold in general, and capacity losses emerge. The stability and control activation analysis provides essential insights into the structural limitations of Max-Pressure control under practical conditions, and serves as a stepping stone for extending the stability analysis to complex urban traffic networks.

The chamber pressure in shield tunnelling is bounded by the lower-limit collapse pressure and the upper-limit blowout pressure, and its accurate determination is essential for construction safety and efficiency. However, existing approaches either neglect the effects of dynamic cutterhead–soil interaction or fail to account for spatial variability in soil. This study develops a three-dimensional probabilistic large-deformation framework for defining reliability-based chamber-pressure windows. By combining coupled Eulerian–Lagrangian modelling, lognormal random fields, and Monte Carlo simulations, the framework accounts simultaneously for dynamic cutterhead–soil interaction, large deformation, and spatial variability in soil strength, allowing both collapse and blowout limits to be evaluated within a unified design approach. Deterministic analyses show that cutterhead configuration strongly modifies the stress-transfer mechanism at the tunnel face. As the opening ratio increases from 15 % to 50 %, the lower-bound pressure required to prevent collapse increases, while the upper-bound pressure associated with blowout decreases, leading to a progressively narrower chamber-pressure window. Stochastic analyses further show that spatial variability in soil strength reduces the stability margin because failure preferentially localizes through weak zones near the tunnel face. Based on these results, a reliability-based procedure is proposed to calibrate collapse and blowout pressure limits for a prescribed target probability of failure. Comparison with chamber-pressure and settlement records from the Qianjiang Tunnel project indicates that field pressure control is governed primarily by collapse risk, and that incorporating both dynamic cutterhead–soil interaction and soil spatial variability provides a more conservative and realistic basis for chamber-pressure design in soft ground.

Incorporation of evolutionary theory into experimental investigations of natural populations was important to the development of plant ecology. We review the prominent role grasses (Poaceae) have played in this historical development. Beginning in the 1920s, the ecotype concept prompted researchers to investigate genetically based differentiation in agronomically important forage grasses using common gardens, first in relation to climatological and biotic factors (grazing and competition), and later in relation to latitude and elevation at collection sites. Adaptive variation was documented for populations in relation to edaphic factors such as soil nutrients and moisture, salinity, serpentine conditions, and heavy metals. Molecular analyses using allozyme and DNA markers show selection favors specific genetic loci in grass populations from different habitats or under changing climatic conditions. Consistent with theoretical expectations for breeding-system effects on genetic structure, highly outcrossing grass species show lower molecular variation among populations and greater variation within populations. In contrast, highly selfing grasses show more genetic variation partitioned among populations and correspndingly reduced variation within populations. Progress has been made in identifying quantitative trait loci important to genetic differentiation in phenotypic traits. Biotic factors such as herbivory, competition and microbial symbionts can act as agents of natural selection and adaptation in grass species. Despite inconsistencies in usage, “ecotype” persists in the literature, but the term has value as evidenced by recent adaptation studies in the Poaceae. Applications of the ecotype concept include reclamation and phytoremediation of sites polluted by metalliferous wastes, restoration of habitats, evolutionary responses to a changing climate, and adaptation of weedy and invasive species.

Joint Embedding Predictive Architectures (JEPAs) offer a compelling framework for learning world models in compact latent spaces, yet existing methods remain fragile, relying on complex multi-term losses, exponential moving averages, pre-trained encoders, or auxiliary supervision to avoid representation collapse. In this work, we introduce LeWorldModel (LeWM), the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings. This reduces tunable loss hyperparameters from six to one compared to the only existing end-to-end alternative. With 15M parameters trainable on a single GPU in a few hours, LeWM plans up to 48x faster than foundation-model-based world models while remaining competitive across diverse 2D and 3D control tasks. Beyond control, we show that LeWM's latent space encodes meaningful physical structure through probing of physical quantities. Surprise evaluation confirms that the model reliably detects physically implausible events.