Evolution-based Shape and Behavior Co-design of Virtual Agents.

We introduce a novel co-design method for autonomous moving agents' shape attributes and locomotion by combining deep reinforcement learning and evolution with user control. Our main inspiration comes from evolution, which has led to wide variability and adaptation in Nature and has significantly improved design and behavior simultaneously. Our method takes an input agent with optional user-defined constraints, such as leg parts that should not evolve or are only within the allowed ranges of changes. It uses physics-based simulation to determine its locomotion and finds a behavior policy for the input design that is used as a baseline for comparison. The agent is randomly modified within the allowed ranges, creating a new generation of several hundred agents. The generation is trained by transferring the previous policy, which significantly speeds up the training. The best-performing agents are selected, and a new generation is formed using their crossover and mutations. The next generations are then trained until satisfactory results are reached. We show a wide variety of evolved agents, and our results show that even with only 10 the overall performance of the evolved agents improves by 50 experiments' performance will improve even more to 150 structures, and it does not require considerable computation resources as it works on a single GPU and provides results by training thousands of agents within 30 minutes.

Current challenges and future of agricultural genomes to phenomes in the USA.

Dramatic improvements in measuring genetic variation across agriculturally relevant populations (genomics) must be matched by improvements in identifying and measuring relevant trait variation in such populations across many environments (phenomics). Identifying the most critical opportunities and challenges in genome to phenome (G2P) research is the focus of this paper. Previously (Genome Biol, 23(1):1-11, 2022), we laid out how Agricultural Genome to Phenome Initiative (AG2PI) will coordinate activities with USA federal government agencies expand public-private partnerships, and engage with external stakeholders to achieve a shared vision of future the AG2PI. Acting on this latter step, AG2PI organized the "Thinking Big: Visualizing the Future of AG2PI" two-day workshop held September 9-10, 2022, in Ames, Iowa, co-hosted with the United State Department of Agriculture's National Institute of Food and Agriculture (USDA NIFA). During the meeting, attendees were asked to use their experience and curiosity to review the current status of agricultural genome to phenome (AG2P) work and envision the future of the AG2P field. The topic summaries composing this paper are distilled from two 1.5-h small group discussions. Challenges and solutions identified across multiple topics at the workshop were explored. We end our discussion with a vision for the future of agricultural progress, identifying two areas of innovation needed: (1) innovate in genetic improvement methods development and evaluation and (2) innovate in agricultural research processes to solve societal problems. To address these needs, we then provide six specific goals that we recommend be implemented immediately in support of advancing AG2P research.

DeepTree: Modeling Trees with Situated Latents.

In this paper, we propose DeepTree, a novel method for modeling trees based on learning developmental rules for branching structures instead of manually defining them. We call our deep neural model "situated latent" because its behavior is determined by the intrinsic state -encoded as a latent space of a deep neural model- and by the extrinsic (environmental) data that is "situated" as the location in the 3D space and on the tree structure. We use a neural network pipeline to train a situated latent space that allows us to locally predict branch growth only based on a single node in the branch graph of a tree model. We use this representation to progressively develop new branch nodes, thereby mimicking the growth process of trees. Starting from a root node, a tree is generated by iteratively querying the neural network on the newly added nodes resulting in the branching structure of the whole tree. Our method enables generating a wide variety of tree shapes without the need to define intricate parameters that control their growth and behavior. Furthermore, we show that the situated latents can also be used to encode the environmental response of tree models, e.g., when trees grow next to obstacles. We validate the effectiveness of our method by measuring the similarity of our tree models and by procedurally generated ones based on a number of established metrics for tree form.

The positivity principle: do positive instructors improve learning from video lectures?

The positivity principle states that people learn better from instructors who display positive emotions rather than negative emotions. In two experiments, students viewed a short video lecture on a statistics topic in which an instructor stood next to a series of slides as she lectured and then they took either an immediate test (Experiment 1) or a delayed test (Experiment 2). In a between-subjects design, students saw an instructor who used her voice, body movement, gesture, facial expression, and eye gaze to display one of four emotions while lecturing: happy (positive/active), content (positive/passive), frustrated (negative/active), or bored (negative/passive). First, learners were able to recognize the emotional tone of the instructor in an instructional video lecture, particularly by more strongly rating a positive instructor as displaying positive emotions and a negative instructor as displaying negative emotions (in Experiments 1 and 2). Second, concerning building a social connection during learning, learners rated a positive instructor as more likely to facilitate learning, more credible, and more engaging than a negative instructor (in Experiments 1 and 2). Third, concerning cognitive engagement during learning, learners reported paying more attention during learning for a positive instructor than a negative instructor (in Experiments 1 and 2). Finally, concerning learning outcome, learners who had a positive instructor scored higher than learners who had a negative instructor on a delayed posttest (Experiment 2) but not an immediate posttest (Experiment 1). Overall, there is evidence for the positivity principle and the cognitive-affective model of e-learning from which it is derived.

3D reconstruction identifies loci linked to variation in angle of individual sorghum leaves.

Selection for yield at high planting density has reshaped the leaf canopy of maize, improving photosynthetic productivity in high density settings. Further optimization of canopy architecture may be possible. However, measuring leaf angles, the widely studied component trait of leaf canopy architecture, by hand is a labor and time intensive process. Here, we use multiple, calibrated, 2D images to reconstruct the 3D geometry of individual sorghum plants using a voxel carving based algorithm. Automatic skeletonization and segmentation of these 3D geometries enable quantification of the angle of each leaf for each plant. The resulting measurements are both heritable and correlated with manually collected leaf angles. This automated and scaleable reconstruction approach was employed to measure leaf-by-leaf angles for a population of 366 sorghum plants at multiple time points, resulting in 971 successful reconstructions and 3,376 leaf angle measurements from individual leaves. A genome wide association study conducted using aggregated leaf angle data identified a known large effect leaf angle gene, several previously identified leaf angle QTL from a sorghum NAM population, and novel signals. Genome wide association studies conducted separately for three individual sorghum leaves identified a number of the same signals, a previously unreported signal shared across multiple leaves, and signals near the sorghum orthologs of two maize genes known to influence leaf angle. Automated measurement of individual leaves and mapping variants associated with leaf angle reduce the barriers to engineering ideal canopy architectures in sorghum and other grain crops.

QuadStack: An Efficient Representation and Direct Rendering of Layered Datasets.

We introduce QuadStack, a novel algorithm for volumetric data compression and direct rendering. Our algorithm exploits the data redundancy often found in layered datasets which are common in science and engineering fields such as geology, biology, mechanical engineering, medicine, etc. QuadStack first compresses the volumetric data into vertical stacks which are then compressed into a quadtree that identifies and represents the layered structures at the internal nodes. The associated data (color, material, density, etc.) and shape of these layer structures are decoupled and encoded independently, leading to high compression rates (4× to 54× of the original voxel model memory footprint in our experiments). We also introduce an algorithm for value retrieving from the QuadStack representation and we show that the access has logarithmic complexity. Because of the fast access, QuadStack is suitable for efficient data representation and direct rendering. We show that our GPU implementation performs comparably in speed with the state-of-the-art algorithms (18-79 MRays/s in our implementation), while maintaining a significantly smaller memory footprint.

PICO: Procedural Iterative Constrained Optimizer for Geometric Modeling.

Procedural modeling has produced amazing results, yet fundamental issues such as controllability and limited user guidance persist. We introduce a novel procedural model called PICO (Procedural Iterative Constrained Optimizer) and PICO-Graph that is the underlying procedural model designed with optimization in mind. The key novelty of PICO is that it enables the exploration of generative designs by combining both user and environmental constraints into a single framework by using optimization without the need to write procedural rules. The PICO-Graph procedural model consists of a set of geometry generating operations and a set of axioms connected in a directed cyclic graph. The forward generation is initiated by a set of axioms that use the connections to send coordinate systems and geometric objects through the PICO-Graph, which in turn generates more objects. This allows for fast generation of complex and varied geometries. Moreover, we combine PICO-Graph with efficient optimization that allows for quick exploration of the generated models and the generation of variants. The user defines the rules, the axioms, and the set of constraints; for example, whether an existing object should be supported by the generated model, whether symmetries exist, whether the object should spin, etc. PICO then generates a class of geometric models and optimizes them so that they fulfill the constraints. The generation and the optimization in our implementation provides interactive user control during model execution providing continuous feedback. For example, the user can sketch the constraints and guide the geometry to meet these specified goals. We show PICO on a variety of examples such as the generation of procedural chairs with multiple supports, generation of support structures for 3D printing, generation of spinning objects, or generation of procedural terrains matching a given input. Our framework could be used as a component in a larger design workflow; its strongest application is in the early rapid ideation and prototyping phases.

Voxel carving-based 3D reconstruction of sorghum identifies genetic determinants of light interception efficiency.

Changes in canopy architecture traits have been shown to contribute to yield increases. Optimizing both light interception and light interception efficiency of agricultural crop canopies will be essential to meeting the growing food needs. Canopy architecture is inherently three-dimensional (3D), but many approaches to measuring canopy architecture component traits treat the canopy as a two-dimensional (2D) structure to make large scale measurement, selective breeding, and gene identification logistically feasible. We develop a high throughput voxel carving strategy to reconstruct 3D representations of sorghum from a small number of RGB photos. Our approach builds on the voxel carving algorithm to allow for fully automatic reconstruction of hundreds of plants. It was employed to generate 3D reconstructions of individual plants within a sorghum association population at the late vegetative stage of development. Light interception parameters estimated from these reconstructions enabled the identification of known and previously unreported loci controlling light interception efficiency in sorghum. The approach is generalizable and scalable, and it enables 3D reconstructions from existing plant high throughput phenotyping datasets. We also propose a set of best practices to increase 3D reconstructions' accuracy.

Multiscale computational models can guide experimentation and targeted measurements for crop improvement.

Computational models of plants have identified gaps in our understanding of biological systems, and have revealed ways to optimize cellular processes or organ-level architecture to increase productivity. Thus, computational models are learning tools that help direct experimentation and measurements. Models are simplifications of complex systems, and often simulate specific processes at single scales (e.g. temporal, spatial, organizational, etc.). Consequently, single-scale models are unable to capture the critical cross-scale interactions that result in emergent properties of the system. In this perspective article, we contend that to accurately predict how a plant will respond in an untested environment, it is necessary to integrate mathematical models across biological scales. Computationally mimicking the flow of biological information from the genome to the phenome is an important step in discovering new experimental strategies to improve crops. A key challenge is to connect models across biological, temporal and computational (e.g. CPU versus GPU) scales, and then to visualize and interpret integrated model outputs. We address this challenge by describing the efforts of the international Crops in silico consortium.

Sculpting Mountains: Interactive Terrain Modeling Based on Subsurface Geology.

Most mountain ranges are formed by the compression and folding of colliding tectonic plates. Subduction of one plate causes large-scale asymmetry while their layered composition (or stratigraphy) explains the multi-scale folded strata observed on real terrains. We introduce a novel interactive modeling technique to generate visually plausible, large scale terrains that capture these phenomena. Our method draws on both geological knowledge for consistency and on sculpting systems for user interaction. The user is provided hands-on control on the shape and motion of tectonic plates, represented using a new geologically-inspired model for the Earth crust. The model captures their volume preserving and complex folding behaviors under collision, causing mountains to grow. It generates a volumetric uplift map representing the growth rate of subsurface layers. Erosion and uplift movement are jointly simulated to generate the terrain. The stratigraphy allows us to render folded strata on eroded cliffs. We validated the usability of our sculpting interface through a user study, and compare the visual consistency of the earth crust model with geological simulation results and real terrains.

Crops In Silico: Generating Virtual Crops Using an Integrative and Multi-scale Modeling Platform.

Multi-scale models can facilitate whole plant simulations by linking gene networks, protein synthesis, metabolic pathways, physiology, and growth. Whole plant models can be further integrated with ecosystem, weather, and climate models to predict how various interactions respond to environmental perturbations. These models have the potential to fill in missing mechanistic details and generate new hypotheses to prioritize directed engineering efforts. Outcomes will potentially accelerate improvement of crop yield, sustainability, and increase future food security. It is time for a paradigm shift in plant modeling, from largely isolated efforts to a connected community that takes advantage of advances in high performance computing and mechanistic understanding of plant processes. Tools for guiding future crop breeding and engineering, understanding the implications of discoveries at the molecular level for whole plant behavior, and improved prediction of plant and ecosystem responses to the environment are urgently needed. The purpose of this perspective is to introduce Crops in silico (cropsinsilico.org), an integrative and multi-scale modeling platform, as one solution that combines isolated modeling efforts toward the generation of virtual crops, which is open and accessible to the entire plant biology community. The major challenges involved both in the development and deployment of a shared, multi-scale modeling platform, which are summarized in this prospectus, were recently identified during the first Crops in silico Symposium and Workshop.

Error-Bounded and Feature Preserving Surface Remeshing with Minimal Angle Improvement.

Surface remeshing is a key component in many geometry processing applications. The typical goal consists in finding a mesh that is (1) geometrically faithful to the original geometry, (2) as coarse as possible to obtain a low-complexity representation and (3) free of bad elements that would hamper the desired application (e.g., the minimum interior angle is above an application-dependent threshold). Our algorithm is designed to address all three optimization goals simultaneously by targeting prescribed bounds on approximation error , minimal interior angle and maximum mesh complexity (number of vertices). The approximation error bound is a hard constraint, while the other two criteria are modeled as optimization goals to guarantee feasibility. Our optimization framework applies carefully prioritized local operators in order to greedily search for the coarsest mesh with minimal interior angle above and approximation error bounded by . Fast runtime is enabled by a local approximation error estimation, while implicit feature preservation is obtained by specifically designed vertex relocation operators. Experiments show that for reasonable angle bounds ( ) our approach delivers high-quality meshes with implicitly preserved features (no tagging required) and better balances between geometric fidelity, mesh complexity and element quality than the state-of-the-art.

FOVEA: a new program to standardize the measurement of foveal pit morphology.

The fovea is one of the most studied retinal specializations in vertebrates, which consists of an invagination of the retinal tissue with high packing of cone photoreceptors, leading to high visual resolution. Between species, foveae differ morphologically in the depth and width of the foveal pit and the steepness of the foveal walls, which could influence visual perception. However, there is no standardized methodology to measure the contour of the foveal pit across species. We present here FOVEA, a program for the quantification of foveal parameters (width, depth, slope of foveal pit) using images from histological cross-sections or optical coherence tomography (OCT). FOVEA is based on a new algorithm to detect the inner retina contour based on the color variation of the image. We evaluated FOVEA by comparing the fovea morphology of two Passerine birds based on histological cross-sections and its performance with data from previously published OCT images. FOVEA detected differences between species and its output was not significantly different from previous estimates using OCT software. FOVEA can be used for comparative studies to better understand the evolution of the fovea morphology in vertebrates as well as for diagnostic purposes in veterinary pathology. FOVEA is freely available for academic use and can be downloaded at: http://estebanfj.bio.purdue.edu/fovea.

A Review of Simulators with Haptic Devices for Medical Training.

Medical procedures often involve the use of the tactile sense to manipulate organs or tissues by using special tools. Doctors require extensive preparation in order to perform them successfully; for example, research shows that a minimum of 750 operations are needed to acquire sufficient experience to perform medical procedures correctly. Haptic devices have become an important training alternative and they have been considered to improve medical training because they let users interact with virtual environments by adding the sense of touch to the simulation. Previous articles in the field state that haptic devices enhance the learning of surgeons compared to current training environments used in medical schools (corpses, animals, or synthetic skin and organs). Consequently, virtual environments use haptic devices to improve realism. The goal of this paper is to provide a state of the art review of recent medical simulators that use haptic devices. In particular we focus on stitching, palpation, dental procedures, endoscopy, laparoscopy, and orthopaedics. These simulators are reviewed and compared from the viewpoint of used technology, the number of degrees of freedom, degrees of force feedback, perceived realism, immersion, and feedback provided to the user. In the conclusion, several observations per area and suggestions for future work are provided.

A flexible pinhole camera model for coherent nonuniform sampling.

The flexible pinhole camera (FPC) allows flexible modulation of the sampling rate over the field of view. The FPC is defined by a viewpoint and a map specifying the sampling locations on the image plane. The map is constructed from known regions of interest with interactive and automatic approaches. The FPC provides inexpensive 3D projection that allows rendering complex datasets quickly, in feed-forward fashion, by projection followed by rasterization. The FPC supports many types of data, including image, height field, geometry, and volume data. The resulting image is a coherent nonuniform sampling (CoNUS) of the dataset that matches the local variation of the dataset's importance. CoNUS images have been successfully implemented for remote visualization, focus-plus-context visualization, and acceleration of expensive rendering effects such as surface geometric detail and specular reflection. A video explaining and demonstrating the FPC is at http://youtu.be/kvFe5XjOPNM.

Quantification of fossil fuel CO2 emissions on the building/street scale for a large U.S. city.

In order to advance the scientific understanding of carbon exchange with the land surface, build an effective carbon monitoring system, and contribute to quantitatively based U.S. climate change policy interests, fine spatial and temporal quantification of fossil fuel CO(2) emissions, the primary greenhouse gas, is essential. Called the "Hestia Project", this research effort is the first to use bottom-up methods to quantify all fossil fuel CO(2) emissions down to the scale of individual buildings, road segments, and industrial/electricity production facilities on an hourly basis for an entire urban landscape. Here, we describe the methods used to quantify the on-site fossil fuel CO(2) emissions across the city of Indianapolis, IN. This effort combines a series of data sets and simulation tools such as a building energy simulation model, traffic data, power production reporting, and local air pollution reporting. The system is general enough to be applied to any large U.S. city and holds tremendous potential as a key component of a carbon-monitoring system in addition to enabling efficient greenhouse gas mitigation and planning. We compare the natural gas component of our fossil fuel CO(2) emissions estimate to consumption data provided by the local gas utility. At the zip code level, we achieve a bias-adjusted Pearson r correlation value of 0.92 (p < 0.001).

Automatic extraction of Manhattan-World building masses from 3D laser range scans.

We propose a novel approach for the reconstruction of urban structures from 3D point clouds with an assumption of Manhattan World (MW) building geometry; i.e., the predominance of three mutually orthogonal directions in the scene. Our approach works in two steps. First, the input points are classified according to the MW assumption into four local shape types: walls, edges, corners, and edge corners. The classified points are organized into a connected set of clusters from which a volume description is extracted. The MW assumption allows us to robustly identify the fundamental shape types, describe the volumes within the bounding box, and reconstruct visible and occluded parts of the sampled structure. We show results of our reconstruction that has been applied to several synthetic and real-world 3D point data sets of various densities and from multiple viewpoints. Our method automatically reconstructs 3D building models from up to 10 million points in 10 to 60 seconds.

Built-in electric field minimization in (In, Ga)N nanoheterostructures.

(In, Ga)N nanostructures show great promise as the basis for next generation LED lighting technology, for they offer the possibility of directly converting electrical energy into light of any visible wavelength without the use of down-converting phosphors. In this paper, three-dimensional computation of the spatial distribution of the mechanical and electrical equilibrium in nanoheterostructures of arbitrary topologies is used to elucidate the complex interactions between geometry, epitaxial strain, remnant polarization, and piezoelectric and dielectric contributions to the self-induced internal electric fields. For a specific geometry-nanorods with pyramidal caps-we demonstrate that by tuning the quantum well to cladding layer thickness ratio, h(w)/h(c), a minimal built-in electric field can be experimentally realized and canceled, in the limit of h(w)/h(c) = 1.28, for large h(c) values.

Visualization of simulated urban spaces: inferring parameterized generation of streets, parcels, and aerial imagery.

Urban simulation models and their visualization are used to help regional planning agencies evaluate alternative transportation investments, land use regulations, and environmental protection policies. Typical urban simulations provide spatially distributed data about number of inhabitants, land prices, traffic, and other variables. In this article, we build on a synergy of urban simulation, urban visualization, and computer graphics to automatically infer an urban layout for any time step of the simulation sequence. In addition to standard visualization tools, our method gathers data of the original street network, parcels, and aerial imagery and uses the available simulation results to infer changes to the original urban layout and produce a new and plausible layout for the simulation results. In contrast with previous work, our approach automatically updates the layout based on changes in the simulation data and thus can scale to a large simulation over many years. The method in this article offers a substantial step forward in building integrated visualization and behavioral simulation systems for use in community visioning, planning, and policy analysis. We demonstrate our method on several real cases using a 200 GB database for a 16,300 km2 area surrounding Seattle.