Real-world Terrain Research Articles

Terrain point cloud inpainting via signal decomposition

The rapid development of 3D acquisition technology has made it possible to obtain point clouds of real-world terrains. However, due to limitations in sensor acquisition technology or specific requirements, point clouds often contain defects such as holes with missing data. Inpainting algorithms are widely used to patch these holes. However, existing traditional inpainting algorithms rely on precise hole boundaries, which limits their ability to handle cases where the boundaries are not well-defined. On the other hand, learning-based completion methods often prioritize reconstructing the entire point cloud instead of solely focusing on hole filling. Based on the fact that real-world terrain exhibits both global smoothness and rich local detail, we propose a novel representation for terrain point clouds. This representation can help to repair the holes without clear boundaries. Specifically, it decomposes terrains into low-frequency and high-frequency components, which are represented by B-spline surfaces and relative height maps respectively. In this way, the terrain point cloud inpainting problem is transformed into a B-spline surface fitting and 2D image inpainting problem. By solving the two problems, the highly complex and irregular holes on the terrain point clouds can be well-filled, which not only satisfies the global terrain undulation but also exhibits rich geometric details. The experimental results also demonstrate the effectiveness of our method.

Walking on Real-world Terrain with an Ankle Exoskeleton in Cerebral Palsy.

Despite medical treatment focused on addressing walking disability, many millions of people with neurological conditions, like cerebral palsy (CP), struggle to maintain independent mobility. Lower limb exoskeletons and exosuits may hold potential for augmenting walking ability. However, it remains unknown whether these wearable robots are safe and beneficial for use outside of highly controlled laboratory environments, the demonstration of which is necessary for clinical translation. Here, we show that a lightweight, portable, ankle exoskeleton with an adaptable one-size-works-for-all assistance controller can improve energy efficiency and walking speed for individuals with CP spanning a wide spectrum of lower limb impairment in a multi-terrain real-world environment. Tested on an outdoor walking route with level, sloped, and stair terrain, robotic assistance resulted in a 15-18% (p = 0.013-0.026) reduction in estimated energy cost and a 7-8% (p = 0.001-0.004) increase in average walking speed across "shorter" 6-minute and "longer" 20-minute walking durations relative to unassisted walking. This study provides evidence that wearable robots may soon improve mobility in neighborhood, school, and community settings for individuals with CP.

A continuous digital elevation representation model for DEM super-resolution

The surface of the Earth is continuous, and obtaining precise elevation data at arbitrary query positions is essential for many applications and analyses. However, existing digital elevation models suffer from a precision gap caused by discretization. Therefore, we develop a new continuous representation model (CDEM) that allows height values to be obtained at any arbitrary query position. Inspired by recent research on the implicit neural representation model, we train an encoder–decoder network to learn CDEM from discrete elevation data for DEM super-resolution tasks. The encoder targets generating latent codes from discrete elevation data, while the decoder composes these latent codes with query positions to predict corresponding elevation values. Such a learning pipeline is well-suited for DEM super-resolution tasks. To improve model accuracy, we also propose predicting the bias of elevation values between the query position and its closest known position. Real-world terrain surfaces exhibit inherent roughness with numerous small variations in localized regions, resulting in high-frequency targets that are difficult for neural networks to fit. In order to facilitate the network’s ability to model data with high-frequency variations, we introduce positional encoding to map query positions into a higher-dimensional space. Compared to the Bicubic interpolation method and state-of-the-art TfaSR model, our method is demonstrated to obtain more accurate elevation values and preserve more details of terrain structure on the TFASR30, Pyrenees, and Tyrol datasets. Specifically, our EBCF-CDEM model demonstrates performance improvements over the TfaSR model, with reductions in the RMSE of elevation, slope, and aspect by 7.03%, 4.81%, and 3.07% respectively at ×4 super-resolution scale on the TFASR30 dataset, by 9.92%, 7.06%, and 6.15% at ×8 super-resolution scale on the Pyrenees dataset. Extensive experiments further validate the generalizability of our EBCF-CDEM, compared to the TfaSR model, our results in terms of RMSE of elevation, slope, and aspect are reduced by 14.78%, 12.12%, and 8.99% respectively at ×8 super-resolution scale on the Tyrol dataset, and 26.92%, 7.14% and 5.36% on the TIFASR30to10 dataset. We release our source code (including the datasets) at https://github.com/AlcibiadesTophetScipio/EBCF-CDEM to reproduce our results and encourage future research.

Reduced-Complexity and Robust Surface Water Path Delineation for Large Real-World Terrain Using Boundary Element Method

Reduced-Complexity and Robust Surface Water Path Delineation for Large Real-World Terrain Using Boundary Element Method

Foot orientation and trajectory variability in locomotion: Effects of real-world terrain.

Capturing human locomotion in nearly any environment or context is becoming increasingly feasible with wearable sensors, giving access to commonly encountered walking conditions. While important in expanding our understanding of locomotor biomechanics, these more variable environments present challenges to identify changes in data due to person-level factors among the varying environment-level factors. Our study examined foot-specific biomechanics while walking on terrain commonly encountered with the goal of understanding the extent to which these variables change due to terrain. We recruited healthy adults to walk at self-selected speeds on stairs, flat ground, and both shallow and steep sloped terrain. A pair of inertial measurement units were embedded in both shoes to capture foot biomechanics while walking. Foot orientation was calculated using a strapdown procedure and foot trajectory was determined by double integrating the linear acceleration. Stance time, swing time, cadence, sagittal and frontal orientations, stride length and width were extracted as discrete variables. These data were compared within-participant and across terrain conditions. The physical constraints of the stairs resulted in shorter stride lengths, less time spent in swing, toe-first foot contact, and higher variability during stair ascent specifically (p<0.05). Stride lengths increased when ascending compared to descending slopes, and the sagittal foot angle at initial contact was greatest in the steep slope descent condition (p<0.05). No differences were found between conditions for horizontal foot angle in midstance (p≥0.067). Our results show that walking on slopes creates differential changes in foot biomechanics depending on whether one is descending or ascending, and stairs require different biomechanics and gait timing than slopes or flat ground. This may be an important factor to consider when making comparisons of real-world walking bouts, as greater proportions of one terrain feature in a data set could create bias in the outcomes. Classifying terrain in unsupervised walking datasets would be helpful to avoid comparing metrics from different walking terrain scenarios.

High-performance adaptive texture streaming for planetary-scale high-mobility information visualization

Planetary-scale terrain requires streaming of world-scale high-resolution data sets in real time. High-performance streaming of world-scale data sets to achieve high mobility in terms of speed and view such as look-around, side-ways, or backward for high fidelity display systems is challenging. Moreover, streaming data sets without any preprocessing overhead require high-performance decompression and re-projection in real time. This work proposes a high-performance adaptive texture streaming system to capture high-speed scenes with view switching capability over a real-world terrain for high-fidelity display system. The system proposes multi-stage pipeline to stream multi-resolution data sets with a stable high-performance rendering rate. The proposed pipeline stages are asynchronously connected using request queues with a cache hierarchy. Efficient algorithms for resource management, prefetching and caching, prioritization of load requests, and hardware-accelerated processing are proposed for texture streaming without frame-rate stuttering. The proposed algorithms adapt to viewer navigation, display resolution, and dynamic execution environment. The proposed method is evaluated for complex scenes, mobility in terms of speed and view, and high-fidelity display. Our method achieved a ground-speed of 2.4x105,1.2x105,and9.0x104 km/h for HD (1024x720), FHD (1920x1080), and UHD (3840x2160) display resolutions respectively with a stable frame rate of 850, 590, and 420 frames per second.

The impact of terrain model source and resolution on snow avalanche modeling

Abstract. Natural hazard models need accurate digital elevation models (DEMs) to simulate mass movements on real-world terrain. A variety of platforms (terrestrial, drones, aerial, satellite) and sensor technologies (photogrammetry, lidar, interferometric synthetic aperture radar) are used to generate DEMs at a range of spatial resolutions with varying accuracy. As the availability of high-resolution DEMs continues to increase and the cost to produce DEMs continues to fall, hazard modelers must often choose which DEM to use for their modeling. We use satellite photogrammetry and topographic lidar to generate high-resolution DEMs and test the sensitivity of the Rapid Mass Movement Simulation (RAMMS) software to the DEM source and spatial resolution when simulating a large and complex snow avalanche along Milford Road in Aotearoa/New Zealand. Holding the RAMMS parameters constant while adjusting the source and spatial resolution of the DEM reveals how differences in terrain representation between the satellite photogrammetry and topographic lidar DEMs (2 m spatial resolution) affect the reliability of the simulation estimates (e.g., maximum core velocity, powder pressure, runout length, final debris pattern). At the same time, coarser representations of the terrain (5 and 15 m spatial resolution) simulate avalanches that run too far and produce a powder cloud that is too large, though with lower maximum impact pressures, compared to the actual event. The complex nature of the alpine terrain in the avalanche path (steep, rough, rock faces, treeless) makes it a suitable location to specifically test the model sensitivity to digital surface models (DSMs) where both ground and above-ground features on the topography are included in the elevation model. Considering the nature of the snowpack in the path (warm, deep with a steep elevation gradient) lying on a bedrock surface and plunging over a cliff, RAMMS performed well in the challenging conditions when using the high-resolution 2 m lidar DSM, with 99 % of the simulated debris volume located in the documented debris area.

Hierarchical and smoothed topographic path planning for large-scale virtual simulation environments

Virtual simulation systems ought to make the most of agents’ realistic behaviors so that users feel immersed in virtual scenarios that mimic the real-world environment. Usually used for training and instruction purposes, agent-based simulations have additional requirements to be considered by the Artificial Intelligence solutions often explored in the development of computer games. To provide a valuable impression of how the simulated agents navigate in the real-world terrain, one of the key requirements indicates that topographic terrain characteristics need to be considered in the path planning tasks. It means that important aspects of the agents’ navigation capabilities have to be properly handled by the used pathfinding algorithms. This problem is even more challenging when large-scale terrain scenarios are considered in state-of-the-art simulation systems as the path planning may be too time-consuming when handling terrain representations with large search spaces. To approach the challenge of computing safe and realistic path solutions for agents inserted in simulation-based learning environments, this work discusses a hierarchical pathfinding algorithm named HPATheta*, showing how to compute terrain relief-aware smoothed paths in reasonable computing time when dealing with path planning in large-scale real-world virtual terrains. Using different simulated real-world virtual terrains, the proposal is validated by performing an extensive experimental campaign with the use of large-scale virtual terrain representations. The experimental results indicate that the proposed approach is able to efficiently address the targeted problem, showing better results compared to other similar path planning algorithms.

Example-based terrain synthesis with pit removal

Artificial terrain synthesis is challenging because natural terrain contains many types of features generated by a diverse range of natural processes over vastly different time-scales. Example-based terrain synthesis has been used to overcome the challenge by using real-world terrain as the basis for synthesis. This produces good local behaviour but can have poor global behaviour. We introduce fluid-flow solutions that can be overlaid on multi-resolution example-based terrain synthesis to improve global realism. We apply these to an existing pixel-based terrain synthesis method and to a novel terrain synthesis method, terrain optimisation, based on texture optimisation. Our fluid-flow solutions improve results for artificial terrains.

Robust Trajectory Optimization Over Uncertain Terrain With Stochastic Complementarity

Trajectory optimization with contact-rich behaviors has recently gained attention for generating diverse locomotion behaviors without pre-specified ground contact sequences. However, these approaches rely on precise models of robot dynamics and the terrain and are susceptible to uncertainty. Recent works have attempted to handle uncertainties in the system model, but few have investigated uncertainty in contact dynamics. In this letter, we model uncertainty stemming from the terrain and design corresponding risk-sensitive objectives for contact-implicit trajectory optimization. In particular, we parameterize uncertainties from the terrain contact distance and friction coefficients using probability distributions and propose a corresponding expected residual minimization cost approach. We evaluate our method in three simple robotic examples, including a legged hopping robot, and we benchmark one of our examples in simulation against a robust worst-case solution. We show that our risk-sensitive method produces contact-averse trajectories that are robust to terrain perturbations. Moreover, we demonstrate that the resulting trajectories converge to those generated by a traditional, non-robust method as the terrain model becomes more certain. Our study marks an important step towards a fully robust, contact-implicit approach suitable for deploying robots on real-world terrain.

Antarctic Water Masses and Ice Shelves: Visualizing the Physics.

High-resolution simulation of global climate physics enables us to model how the climate may change under a variety of future scenarios. Such simulations produce vast amounts of information and dense datasets. If interrogated in tandem, these datasets can provide holistic, vital information on Earth's many integrated systems by revealing the manifold interrelated properties of the atmosphere, ocean, and polar ice, framed by real-world terrain in three-dimensional space as they vary over time. To accomplish this, climate scientists have joined with computer scientists and an artist to develop techniques enabling scientists to see these relationships. The impact of ocean water properties on Antarctic ice shelves illustrates the benefit of this analysis in understanding land ice melt rates and thus sea-level rise.

Extracting and visualising glacial ice flow directions from Digital Elevation Models using greyscale thinning and directional trend analyses

Flow pattern reconstructions for past glaciations are based on the analysis of the spatial distribution of subglacial landforms, and streamlined subglacial landforms (oriented parallel or sub-parallel to the ice flow) are regarded as the main indicators of the previous ice flow. Manual mapping of these landforms is a time consuming and subjective process, making semi-automated mapping (SAM) methods attractive. We present a novel SAM method that extracts mean directional trends from a digital elevation model (DEM) by performing greyscale thinning on its derived slope raster, and produce directional statistics from the lines extracted from the resulting skeleton.The method has been tested on artificial (synthetically generated) surfaces to demonstrate its potential for detecting directions of idealized landforms. The tests carried out on the artificial surfaces reveal that a landform feature has to be at least six (raster) cells wide for its direction to be properly detected by the method. The application of the method on real-world terrain data shows that the method is capable of reconstructing glacial flow patterns, and that it is robust to reasonable variations in the method parameters. We believe that the method should also have great potential for detecting directional trends in the general field of geomorphology. The user friendliness of the method and the simple interpretation of the results should make this a useful tool for geomorphological studies.

Virtual testbed for monocular visual navigation of small unmanned aircraft systems

Monocular visual navigation methods have seen significant advances in the last decade, recently producing several real-time solutions for autonomously navigating small unmanned aircraft systems without relying on the Global Positioning System (GPS). This is critical for military operations that may involve environments where GPS signals are degraded or denied. However, testing and comparing visual navigation algorithms remains a challenge since visual data is expensive to gather. Conducting flight tests in a virtual environment is an attractive solution prior to committing to outdoor testing. This work presents a virtual testbed for conducting simulated flight tests over real-world terrain and analyzing the real-time performance of visual navigation algorithms at 31 Hz. This tool was created to ultimately find a visual odometry algorithm appropriate for further GPS-denied navigation research on fixed-wing aircraft, even though all of the algorithms were designed for other modalities. This testbed was used to evaluate three current state-of-the-art, open-source monocular visual odometry algorithms on a fixed-wing platform: Direct Sparse Odometry, Semi-Direct Visual Odometry, and ORB-SLAM2 (with loop closures disabled).

Development of Scaled Real Time Augmented Reality Sandbox

This paper presents a scaled version of Augmented Reality Sandbox (ARSAND). Original ARSAND consist of short throw projector, Kinect 3D camera, dedicated computer with high end graphic card and sandbox. ARSAND has been gaining recognition and popularity for its rapid and accurate capabilities in designing landscapes and simulating geographical real-world terrain. The proposed scaled version of the ARSAND consists of ultra-short throw projector with different configuration, Kinect 3D camera, dedicated computer with high end graphic card, and calibrated sandbox with less amount of sand. The proposed scaled version of ARSAND is able to perform basic functions of the original ARSAND such as simulating real-world terrain and rain fall.

Coherent multi-layer landscape synthesis

We present an efficient method for generating coherent multi-layer landscapes. We use a dictionary built from exemplars to synthesize high-resolution fully featured terrains from input low-resolution elevation data. Our example-based method consists in analyzing real-world terrain examples and learning the procedural rules directly from these inputs. We take into account not only the elevation of the terrain, but also additional layers such as the slope, orientation, drainage area, the density and distribution of vegetation, and the soil type. By increasing the variety of terrain exemplars, our method allows the user to synthesize and control different types of landscapes and biomes, such as temperate or rain forests, arid deserts and mountains.

The functional coupling of gaze and gait when walking over real-world rough terrain

The functional coupling of gaze and gait when walking over real-world rough terrain

Simplification of real-world terrain for outdoor acoustic propagation modeling

Approaches to outdoor sound propagation include models of attenuation for a set of standardized barriers and terrains. These various approaches take care to represent the physical nature of each barrier type. However selection of the appropriate barrier model for a real-world terrain segment is rarely presented. Complex logic is required to determine the appropriate barrier model to apply in the instances where the selection is addressed. This paper presents the Critical Point Analysis Terrain Simplification (CPATS), an algorithm for reducing the complexity of real-world terrain profiles. CPATS implements a finite differencing schema to determine critical points within the terrain segment. The results of this model are compared to a maximum deviation algorithm (MaxDev) of reducing terrain complexity. A statistical comparison between CPATS and MaxDev terrains is completed showing the superior capabilities of the CPATS model. The effect of the different simplified terrains is examined using the outdoor sound propagation code NORD2000.

Evaluating positional accuracy of Pleiades 1A satellite imagery in exploiting foreign natural resources

In this research, we’ve carried out a geo-positioning accuracy test for the unaccessable area using Pleiades 1A satellite image which was captured in 2014 for the city of Yakutsk, Russia with 50 cm spatial resolution. The research area has been considered as a candidate site for the oversea resources development and on-site field surveying of the area has not been possible so far. There were two difficulties to use Pleiades 1A for the research area. First, we had to use inaccurate ground control points due to inaccessibility of research area. Second, there were huge differences between real-world terrain of Yakutsk and the calculated height range by the normalized offset and scale parameters of the vendor-provided rational polynomial coefficients (RPCs). We’ve tried to enhance geo-positioning accuracy by sorting out these two difficulties and finally tried to evaluate the usability of Pleiades 1A for the unaccessible area. We’ve applied the direct method for RPCs refinement by using control points captured from Google Earth and ASTER Global Digital Elevation Models (GDEM). As a result of this method, we were able to improve the vertical-accuracy significantly. However we also found that this method didn’t improve the horizontal-accuracy, because of low accuracy of reference points from Google Earth or GDEM. It is required to carry out additional research to improve the geo-positioning accuracy of Pleiades 1A image for the unaccessible area.

Eye, head, and foot tracking during locomation over real-world complex terrain

Successful locomotion over complex terrain such as a rocky trail requires walkers to place each step on a safe foothold with high spatiotemporal precision while also taking the future terrain into account. We have previously shown that humans guide foot placement using a visual control strategy that exploits the physical dynamics of the bipedal gait cycle (e.g. Matthis, Barton & Fajen, 2014). However, this work was limited to locomotion over virtual terrain projected onto flat surfaces within the laboratory. In the current project, we develop a research protocol to allow accurate and integrated tracking of the eye, head, and feet during locomotion over real-world complex and rocky terrain, a protocol that has been previously limited by numerous technical challenges. This protocol allows us to test predictions derived from the results of previous lab-based studies on the visual control of foot placement in real-world terrains, and opens new opportunities to study the visual control of locomotion outside a laboratory setting. Our protocol utilizes a Positive Science mobile eye tracker with a GoPro scene camera and inertial measurement units (IMUs) attached to the head, trunk, and feet. Each IMU integrates the output of temperature-calibrated, tri-axial accelerometers, gyroscopes, and magnetometers to determine the sensor's orientation within a world-centered reference frame. The orientation quaternions produced by the head and trunk-mounted IMUs were used to rotate the gaze vectors recorded by the eye tracker in order to specify gaze fixations within a body-centered reference frame and relate gaze location to foot placement. In a pilot study, subjects walked across different types of real-world terrain while performing different distraction tasks. Preliminary results reveal that foot placement in complex terrain is highly demanding of attentional resources, and that the control strategies used to guide foot placement vary dramatically according to the difficulty of the terrain being traversed. Meeting abstract presented at VSS 2015. Language: en

Terradynamically streamlined shapes in animals and robots enhance traversability through densely cluttered terrain

Many animals, modern aircraft, and underwater vehicles use fusiform, streamlined body shapes that reduce fluid dynamic drag to achieve fast and effective locomotion in air and water. Similarly, numerous small terrestrial animals move through cluttered terrain where three-dimensional, multi-component obstacles like grass, shrubs, vines, and leaf litter also resist motion, but it is unknown whether their body shape plays a major role in traversal. Few ground vehicles or terrestrial robots have used body shape to more effectively traverse environments such as cluttered terrain. Here, we challenged forest-floor-dwelling discoid cockroaches (Blaberus discoidalis) possessing a thin, rounded body to traverse tall, narrowly spaced, vertical, grass-like compliant beams. Animals displayed high traversal performance (79 ± 12% probability and 3.4 ± 0.7 s time). Although we observed diverse obstacle traversal strategies, cockroaches primarily (48 ± 9% probability) used a novel roll maneuver, a form of natural parkour, allowing them to rapidly traverse obstacle gaps narrower than half their body width (2.0 ± 0.5 s traversal time). Reduction of body roundness by addition of artificial shells nearly inhibited roll maneuvers and decreased traversal performance. Inspired by this discovery, we added a thin, rounded exoskeletal shell to a legged robot with a nearly cuboidal body, common to many existing terrestrial robots. Without adding sensory feedback or changing the open-loop control, the rounded shell enabled the robot to traverse beam obstacles with gaps narrower than shell width via body roll. Such terradynamically ‘streamlined’ shapes can reduce terrain resistance and enhance traversability by assisting effective body reorientation via distributed mechanical feedback. Our findings highlight the need to consider body shape to improve robot mobility in real-world terrain often filled with clutter, and to develop better locomotor-ground contact models to understand interaction with 3D, multi-component terrain.