Background, Southwest Research Institute® has developed a wide range of traffic management solutions for state and local governments across the country. The high human and economic toll of traffic accidents has made their mitigation an area of particular interest. Recent advances in machine learning (ML) and data-source availability have enabled the prediction of traffic accident probabilities, allowing…, Approach, First, roadway data was gathered for the entire state of Texas and static features were extracted before information-preserving downsampling was performed. Next, dynamic data with one-hour temporal resolution was collected from a variety of sources, including weather, speed, and crowdsourced incident reports, covering the past year. The combined static and dynamic data were used as features in…, Accomplishments, Three ML models were developed to predict the likelihood of traffic accidents from the synthesis of spatiotemporal data. First, a static model – implemented as a message passing neural network (MPNN) – was developed using only roadway features and connectivity. After the development of the static model, two dynamic models were produced. Due to difficulties in data collection, the speed…
Type: IRD Synopsis
Background, This research aims to answer two open questions within the space of creating lidar maps using vehicle mounted sensors: how to automatically incorporate expert human knowledge into machine learning models to remove objects in the scene that don’t belong in the map; and how to execute those models in real time on vehicle-relevant hardware rather than the typical high-powered desktop hardware found…, Approach, The first step to addressing these challenges is to design a structured representation of the data which can be fed into a neural network in such a way to take advantage of current neural network research. The most common output of a lidar sensor is a cloud of points in 3D space. While this format is easy to transport and visualize, it discards a lot of the spatial information between points. Our…, Accomplishments, The team has designed and tested a neural network architecture to perform the map reduction and creation task. We have also adapted relevant data sources from both publicly available datasets and local data collection, to make them compatible with our neural network design, significantly expanding the available training data without compromising the network’s capability or efficiency. Finally,…
Type: IRD Synopsis
Background, Traditional deep learning models are known to be biased black boxes, due to training data diversity and chosen complexity of the model. Therefore, the ability for models to produce explainable, human-interpretable output is becoming increasingly desired and explored, especially in computer vision tasks. This research focused on a component of explainability: Uncertainty Quantification (UQ)…, Approach, The research had two main objectives. First, UQ was implemented with a baseline semantic segmentation model to prove validity. This was accomplished by utilizing Monte Carlo dropout to simulate Bayesian inference, as seen in Figure 1. By introducing dropout layers to a SegNet deep learning model and then passing in the same input multiple times, the network now outputs a distribution of output…, Accomplishments, These efforts resulted in two semantic segmentation networks capable of communicating levels of uncertainty in their respective outputs. These uncertainty maps can be compared in Figure 2. After conducting proper model validation on both models, the researchers were able to determine that not only does the UNet model have a higher prediction accuracy than the SegNet model, but the UNet model also…
Type: IRD Synopsis
Background, Over the past few years, SwRI's Powertrain Engineering Division and Intelligent Systems Division have collaborated on a variety of Connected and Automated Vehicle (CAV) technologies that improve efficiency ranging from the vehicle level [Next-Generation Energy Technologies for Connected and Automated On-Road Vehicles (NEXTCAR)] to the intersection level (Energy Efficient Mobility Systems) up to…, Approach, Phase I of the program focused on leveraging instrumented Class 8 trucks and demonstrating the potential fuel economy benefits of CAV technologies akin to the NEXTCAR program. The team refined the traffic simulator setup built for NEXTCAR/Energy Efficient Mobility Systems (EEMS) program and generated additional scenarios as needed to represent a good mix or urban, suburban and highway driving.…, Accomplishments, Phase 1 testing with the Class 8 trucks was successful and resulted in significant improvements in fuel economy as showing in the following tables: Table 1: Results for modified NREL Port Drayage cycle, Test Result, Nominal, Confidence Interval, Fuel Saved 14.24% ± 1.81% Improvement 16.61% ± 2.11% Table 2: Results for traffic simulation generated traces, Test Result, Nominal, Confidence Interval, Fuel Saved 6.83% ± 4.61% Improvement 7.34% ± 4.95% Phase 2 is currently being conducted and results are not yet available. The Ego Vehicle interacting with traffic in the simulator Figure 1: The Ego Vehicle (red) interacting with traffic in the simulator, the Development Vehicle mounted on the Hub Dynamometer. Graph showing quantify benefits of NEXTCAR style technologies on Class VIII Trucks…
Type: IRD Synopsis
Background, The objective of this internal research project was to create a robust and automated approach for optimizing object-detecting deep neural networks (DNN) for new computing platforms for which the DNN will need to be deployed. Having the ability to optimize a DNN quickly and easily for new platforms is essential for successful long-term commercial deployment of any system that utilizes DNN based…, Approach, To achieve an approach for automating the search of efficient DNNs, we had to first curate a clean dataset that could be used as an automated metric for our search algorithm. The data used to train the original Active-Vision object detection network was found to have deficiencies, mainly regarding lack of data collected in urban environments such as city intersections, which caused the algorithm…, Accomplishments, This research resulted in various model topologies based on the yoloV5 framework that can run at or faster than the desired five frames per second on the different hardware platforms we were targeting. The techniques used to achieve these speeds are generic enough to allow for exploration of models on different platforms and will help enable Active-Vision and other object detection computer…
Type: IRD Synopsis
Background, Many future space missions will require robots with high levels of autonomy that can interact with objects in space, just as current terrestrial robots operate. This requires reasoning about and operating in three-dimensional (3D) environments, which in turn requires an autonomous robot to reconstruct 3D models of objects in its environment. However, these space robots will face severe…, Approach, We adapted the Fast Lightweight Mesh Estimation Algorithm (FLaME) for use in space applications. This algorithm using variational smoothing on Delaunay graphs to reconstruct 3D surfaces in real time on limited hardware [1, 2]. Delaunay triangulation decomposes an image into geometric primitives, and then a regularization function aligns the vertices of the primitives into a mesh that is…, Accomplishments, We adopted FLaME for space applications. We built a pipeline with space-specific preprocessing steps to eliminate background objects from the reconstruction, and we modified the algorithm to work with long range views of target objects. Furthermore, we ported the algorithm to hardware suitable for space applications.
Type: IRD Synopsis
Background, This research enables robotics in space by leveraging existing SwRI robotics capabilities and transforming them to meet the current and future needs of the space industry. The space industry is pushing for industrial-style robotics capabilities to enable in-space servicing, repair, and maintenance (ISAM). While there have been a few early missions in this arena, the industry needs robotics for…, Approach, This project advances the technology for autonomous resident space object (RSO) characterization, on-orbit refueling, and ISAM through advancing and adapting terrestrial vision systems and path planning., Vision systems in space:, Equip cameras and other sensing modalities with machine vision. This is accomplished by using low-power field-programmable gate arrays (FPGAs) to enable passive advanced image processing in space for building a 3D model of an RSO and identifying rotation., Dynamic path planning:, Use hardware/software to simulate planning robot paths while considering the variables of space operations. Standard robotic arms and a power-constrained processor will complete tasks while minimizing induced momentum., Accomplishments, Before moving to hardware in future phases of this effort, our team is starting development of motion planning algorithms in a simulation environment called Drake. The goal of the simulation component is to both simulate, plan, and command motions of a robotic arm in microgravity that minimize momentum imparted to the satellite base. We have created a Python software package, based on Drake, that…
Type: IRD Synopsis
Background, Underwater robotics is an emerging application of autonomous vehicles. SwRI has extensive experience in designing autonomy software for ground and aerial vehicles but has not yet developed underwater autonomy. One of the major open research topics for underwater autonomy is high-resolution mapping and localization. In ground and aerial domains, SwRI has solved this problem with state-of-the-art…, Approach, A simulated ROV was designed in Gazebo with a 9-degree-of-freedom inertial measurement unit (IMU) and multiple 3D FLIS that mimic real-world sensors. The ROV traveled in simulated environments under user control and interfaced with different publicly available SLAM algorithms using the Robot Operating System (ROS) to determine position and to generate environment maps. Several simulated…, Accomplishments, This work produced a recommendation for which SLAM algorithms are most effective in various environmental conditions and sensor configurations. It determined which sensor configurations balance field of view and rate of return so that the FLIS is usable by existing SLAM algorithms. It was found that no SLAM algorithm achieves sufficiently accurate localization in all environments considered here…
Type: IRD Synopsis
Background, Alzheimer’s disease (AD) is an irreversible, progressive neurological brain disorder that slowly destroys the memory and thinking skills in people, and, eventually, their ability to carry out the simplest tasks. Estimates vary, but experts suggest that more than 5.5 million Americans, most of them age 65 or older, may have dementia caused by Alzheimer’s. Currently there is no known cure or…, Approach, The approach for this project is to formulate drug-loaded, stable, blood-brain barrier (BBB) targeting liposome formulations and to develop experimental in-vitro models and computational models that can precisely evaluate the delivery of those liposome-nanoparticles based on the interaction of the BBB and liposome-nanoparticles., Accomplishments, The following project metrics are briefly summarized here: Goal: Prepare stable, nano-sized liposomes from biocompatible materials. Metric: Formulations with submicron sizes that remain size stable, +/- 10 %, after 2 weeks. Level of Success: COMPLETED- <200 nm sized liposomes were prepared, and the particle size remained stable under refrigeration for at least 2 weeks. Goal: Prepare…,
Type: IRD Synopsis
Background, There is a growing recognition that a large proportion of SARS-CoV-2-infected individuals continue to experience a broad range of symptoms after recovering from the initial bout of the COVID-19 illness. These patients are colloquially referred to as “COVID long haulers,” and the illness as “Long COVID.” There are over 35 million COVID survivors in U.S. and 3.5 million estimated in Texas. Long…, Approach, This project is a collaborative effort between Southwest Research Institute, UT Health San Antonio (UTHSA), and UT San Antonio. The project will use the UTHSA COVID-19 Infectious Diseases Outpatient Clinic cohort (>12,000 patients diagnosed with acute COVID-19) by systematic deep phenotyping of clinical characteristics from different data sources (data warehouse, clinical chart review,…, Accomplishments, The project is still in its early stages and the collaborating institutions are in the process of onboarding staff.
Type: IRD Synopsis
