Explore the vast range of titles available.

Discretely monitoring traffic systems and tracking payloads on vehicle targets can be challenging when traversal occurs off main roads where overhead traffic cameras are not present. This work proposes a portable roadside vehicle detection system as part of a solution for tracking traffic along any path. Training semantic segmentation networks to automatically detect specific types of vehicles while ignoring others will allow the user to track payloads present only on certain vehicles of interest, such as train cars or semi-trucks. Different vision sensors offer varying advantages for detecting targets in changing environments and weather conditions. To analyze the benefits of both, corresponding LiDAR and camera data were collected at multiple roadside sites and then trained on separate semantic segmentation networks for object detection. A custom CNN architecture was built to handle highly asymmetric LiDAR data, and a network inspired by DeepLabV3+ was used for camera data. The performance of both networks was evaluated, and showed comparable accuracy. Inferences run on embedded platforms showed real-time execution matching the performance on the training hardware for edge deployments anywhere. Both camera and LiDAR semantic segmentation networks were successful in identifying vehicles of interest from the proposed viewpoint. These highly accurate vehicle detection networks can pair with a tracking mechanism to establish a non-intrusive roadside detection system.

Simulation has proven to be a highly effective tool for validating autonomous systems while lowering cost and increasing safety. Currently, several dedicated simulation environments exist, but they are limited in terms of environment size, visual quality, and feature sets. As a result, many researchers have begun to consider repurposing game engines as simulators to take advantage of their greater flexibility, scalability, and graphical fidelity. This study investigates a robotics simulation environment based on the Unity3D game engine and Robot Operating System (ROS) middleware, collectively referred to as ROS-Unity3D, and compares it to the popular ROS-Gazebo robotics simulation environment. They are compared in terms of their architecture, environment creation process, resource usage, and accuracy while simulating an autonomous ground robot in real-time. Overall, with its variety of supported file types and powerful scripting interface for creating custom functionality, ROS-Unity3D is found to be a viable alternative to ROS-Gazebo. Test results indicate that ROS-Unity3D scales better to larger environments, has higher shadow quality, achieves the same or better visual-based SLAM performance, and is more capable of real-time LiDAR simulation than ROS-Gazebo. As for its advantages over ROS-Unity3D, ROS-Gazebo has a more streamlined interface between ROS and Gazebo, has more existing sensor plugins, and is more computer resource efficient for simulating small environments.

Existing lidar-based semantic segmentation algorithms and datasets focus on autonomous vehicles operating in urban environments. This has greatly improved the safety and reliability of these autonomous vehicles in predictable scenery. A new dataset provides lidar data focusing on off-road environments as seen by autonomous ground vehicles, ushering in a new era of off-road exploration capabilities. To the best of our knowledge, no new algorithms have been developed specifically for this unstructured environment. To gain an understanding of how existing algorithms perform in an off-road environment, we assess the baseline performance of four algorithms, KPConv, SalsaNext, Cylinder3D, and SphereFormer, on a commonly used on-road dataset, SemanticKITTI. We then compare the results with an off-road dataset, RELLIS-3D. We discuss the degradation of each algorithm on the off-road dataset and investigate potential causes such as class imbalance, inconsistencies in the labeled data, and the inherent difficulty of segmenting off-road environments. We present the strengths and weaknesses of each algorithm’s segmentation abilities and provide a comparison of the runtime of each algorithm for real-time capabilities. This is crucial for identifying what network architecture features are potentially the most beneficial for unstructured scenes. A robust, open-source software implementation via docker containers and bash scripts provides simple, repeatable execution of all algorithm training and evaluations. All code is publicly available at https://github.com/UA-Lidar-Segmentation-Research.

Existing lidar-based semantic segmentation algorithms and datasets focus on autonomous vehicles operating in urban environments. This has greatly improved the safety and reliability of these autonomous vehicles in predictable scenery. A new dataset provides lidar data focusing on off-road environments as seen by autonomous ground vehicles, ushering in a new era of off-road exploration capabilities. To the best of our knowledge, no new algorithms have been developed specifically for this unstructured environment. To gain an understanding of how existing algorithms perform in an off-road environment, we assess the baseline performance of four algorithms, KPConv, SalsaNext, Cylinder3D, and SphereFormer, on a commonly used on-road dataset, SemanticKITTI. We then compare the results with an off-road dataset, RELLIS-3D. We discuss the degradation of each algorithm on the off-road dataset and investigate potential causes such as class imbalance, inconsistencies in the labeled data, and the inherent difficulty of segmenting off-road environments. We present the strengths and weaknesses of each algorithm’s segmentation abilities and provide a comparison of the runtime of each algorithm for real-time capabilities. This is crucial for identifying what network architecture features are potentially the most beneficial for unstructured scenes. A robust, open-source software implementation via docker containers and bash scripts provides simple, repeatable execution of all algorithm training and evaluations. All code is publicly available at https://github.com/UA-Lidar-Segmentation-Research .

In this paper we present a novel hardware architecture for real-time image compression implementing a fast, searchless iterated function system (SIFS) fractal coding method. In the proposed method and corresponding hardware architecture, domain blocks are fixed to a spatially neighboring area of range blocks in a manner similar to that given by Furao and Hasegawa. A quadtree structure, covering from 32 × 32 blocks down to 2 × 2 blocks, and even to single pixels, is used for partitioning. Coding of 2 × 2 blocks and single pixels is unique among current fractal coders. The hardware architecture contains units for domain construction, zig-zag transforms, range and domain mean computation, and a parallel domain-range match capable of concurrently generating a fractal code for all quadtree levels. With this efficient, parallel hardware architecture, the fractal encoding speed is improved dramatically. Additionally, attained compression performance remains comparable to traditional search-based and other searchless methods. Experimental results, with the proposed hardware architecture implemented on an Altera APEX20K FPGA, show that the fractal encoder can encode a 512 × 512 × 8 image in approximately 8.36 ms operating at 32.05 MHz. Therefore, this architecture is seen as a feasible solution to real-time fractal image compression.

Retention and graduation rates for engineering disciplines are significantly lower than desired, and research literature offers many possible causes. Engineering learning communities provide the opportunity to study relationships among specific causes and to develop and evaluate activities designed to lessen their impact. This paper details an engineering learning community created to combat three common threats to academic success of engineering students: financial difficulties, math deficiencies, and the lack of a supportive engineering culture. The project tracks participants in the learning community from first year through graduation to assess the effectiveness of its activities in improving retention and graduation rates. Scholarships were made available to address the financial difficulties; tutors, mentors, study groups, and a “freshman-to-sophomore bridge” summer program were provided to address math deficiencies; cohort engineering courses, active learning techniques, required group meetings, required group study sessions, dedicated study space, and dedicated faculty advisors were used to promote a sense of community. Quantitative retention and graduation rates for the cohort are compared to other engineering groups at the same institution. Qualitative results collected via student surveys and interviews, and lessons learned by project administrators are also presented. Retention and graduation rates of the cohort are better than those of comparable groups at the same institution. Graduation rates based upon freshman math placement are also higher than comparable groups. 

High-level design tools for the design and specification of avionic systems and real-time systems currently exist. However, real-time, hardware-in-the-loop simulations of avionic systems are based upon principles fundamentally different than those used to design avionic systems and represent a specialized case of real-time systems. As a result, the high-level software tools used to design avionic systems and real-time systems cannot be applied to the design of real-time, hardware-in-the-loop simulations of avionic systems. For this reason, such simulations of avionic systems should not be considered part of the domain containing avionic systems or general-purpose real-time systems and should be considered as an application domain unto itself for which design tools are unavailable. To fill this void, this dissertation proposes a framework for the design and specification of real-time, hardware-in-the-loop simulations of avionic systems. This framework is based upon a new specification language called the Simulation Architecture Description Language. This specification language is a graphical language with constructs and semantics defined to provide the user with the capability to completely define the simulation and its software execution characteristics at various levels of abstraction. The language includes a new method for combining precedence constraints for a single software process. These semantics provide a more accurate description of the behavior of software systems having a dynamic job structure than existing semantics. An environment that supports the execution of simulation software having the semantics defined within this language is also described. A toolset that interfaces to the language and provides additional functionality such as design analysis, schedulability analysis, and simulation file generation is also discussed. This framework provides a complete design and specification environment for real-time, hardware-in-the-loop simulations of avionic systems.

Excavation will hold a key role for future lunar missions. NASA has stated that \"advances in lunar regolith mining have the potential to significantly contribute to our nation's space vision and NASA space exploration operations.\" [1]. The Lunabotics Mining Competition is an event hosted by NASA that is meant to encourage \"the development of innovative lunar excavation concepts from universities which may result in clever ideas and solutions which could be applied to an actual lunar excavation device or payload.\" [2]. Teams entering the competition must \"design and build a remote controlled or autonomous excavator, called a lunabot, that can collect and deposit a minimum of 10 kilograms of lunar simulant within 10 minutes.\" [2]. While excavation will play an important part in lunar missions, there will still be many other tasks that would benefit from robotic assistance. An excavator might not be as well suited for these tasks as other types of robots might be. For example a lightweight rover would do well with reconnaissance, and a mobile gripper arm would be fit for manipulation, while an excavator would be comparatively clumsy and slow in both cases. Even within the realm of excavation it would be beneficial to have different types of excavators for different tasks, as there are on Earth. The Alabama Lunabotics Team at the University of Alabama has made it their goal to not only design and build a robot that could compete in the Lunabotics Mining Competition, but would also be a multipurpose tool for future NASA missions. The 2010-2011 resulting robot was named the Modular Omnidirectional Lunar Excavator (MOLE). Using the Systems Engineering process and building off of two years of Lunabotics experience, the 20ll-2012 Alabama Lunabotics team (Team NASACAR) has improved the MOLE 1.0 design and optimized it for the 2012 Lunabotics Competition rules [I]. A CAD model of MOLE 2.0 can be seen below in Fig. 1.

This thesis presents a new, cost effective multiprocessing architecture which can be constructed using off-the-shelf components. The proposed architecture extends the current generation of single-board-computer systems to include a low cost, supplemental interprocessor communication network. The resulting extended single-board-computer multiprocessor (ESBCM) architecture is shown to have improved scalability characteristics and is much more flexible than current multiprocessor designs. It also directly supports many real-time and fault-tolerant constructs not previously supported. Extensive empirical analysis using discrete event simulations and Monte Carlo techniques indicate that the ESBCM architecture will generally outperform standard bus-based multiprocessors.