Off-line Mapping Research Articles

Unsupervised Appearance Map Abstraction for Indoor Visual Place Recognition With Mobile Robots

Visual Place Recognition (VPR), the task of identifying the place where an image has been taken from, is at the core of important robotic problems as relocalization, loop-closure detection or topological navigation. Even for indoors, the focus of this work, VPR is challenging for a number of reasons, including real-time performance when dealing with large image databases ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\sim10^{4}$</tex-math></inline-formula> ) (probably captured by different robots), or the avoidance of Perceptual Aliasing in environments with repetitive structures and scenes. In this letter, we tackle these issues by proposing an off-line mapping technique that abstracts a dense database of georeferenced images without particular order into a Multivariate Gaussian Mixture Model, by creating soft clusters in terms of their similarity in both pose and appearance. This abstract representation is obtained through an Expectation-Maximization algorithm and plays the role of a simplified map. Since querying this map yields a probability of being in a cluster, we exploit this “belief” within a Bayesian filter that regards previous query images and a topological map between clusters to perform more robust VPR. We evaluate our proposal in two different indoor datasets, demonstrating comparable VPR precision to querying the full database while incurring in shorter query times and handling Perceptual Aliasing for sequential navigation.

Offline mapping for autonomous vehicles with low-cost sensors

This paper presents an offline mapping algorithm for autonomous vehicles (AV) with low-cost sensors. The mapping algorithm consists of five key steps. First, data pre-processing is conducted to calibrate the original odometry data. Then, based on a 2D laser scanner and the calibrated odometry data, a virtual 3D light detection and ranging (LiDAR) is built. In the third step, loop closure is performed to search the revisited region and calculate the distance displacement. Afterward, the optimizer is applied to generate the final trajectory. Finally, by fusing the point cloud data from virtual 3D LiDAR and the final trajectory, the point cloud map is generated. Field experiments are conducted in both open and urban areas. In these two cases, satisfactory point cloud maps are built even the vehicle travels a long distance with loop closure, and the constructed point cloud map can be used for AV localization.

Topography analysis using wearable devices and its integration in navigation systems

Automatic navigation in an unknown environment raises various challenges as many cues about orientation are difficult to perceive without the use of vision. Though assisted aids such as GPS devices help in route finding, still it fails to fulfill safety requirements. This paper proposes a framework that provides accurate guiding and information on the route traversal and the topography of the road ahead. The framework is composed of technologies such as Lumigrids, Drone, GPS, Mobile applications, Cloud storage which are used to map the road surface and generate proper navigation guidance to the end user. This is done in three stages; [1]. Off-line mapping of the road surface and storing this information in the cloud. 2. Wearable technology used for obtaining in real-time surface information and comparing it to the data on the cloud facilitating accurate and safer navigation 3. Updating the cloud information with information collected by the pedestrian

Off-line mapping of multi-rate dependent task sets to many-core platforms

This paper presents an approach to execute safety-critical applications on multi- and many-core processors in a predictable manner. We investigate three concrete platforms: the Intel Single-chip Cloud Computer, the Texas Instruments TMS320C6678 and the Tilera TILEmpower-Gx36. We define an execution model to safely execute dependent periodic task sets on these platforms. The four rules of the execution model entail that an off-line mapping of the application to the platform must be computed. The paper details our approach to automatically compute a valid mapping. Furthermore, we evaluate our approach, which is based on constraint programming, by applying it to several task sets that are derived from industrial applications.

Robust Scheduling of Task Graphs under Execution Time Uncertainty

Effective multicore computing requires to make efficient usage of the computational resources on a chip. Offline mapping and scheduling can be applied to improve the performance, but classical approaches require considerable a priori knowledge of the target application. In a practical setting, precise information is often unavailable; one can then resort to approximate time and resource usage figures, but this usually requires to make conservative assumptions. The issue is further stressed if real-time guarantees must be provided. We tackle predictable and efficient nonpreemptive scheduling of multitask applications in the presence of duration uncertainty. Hard real-time guarantees are provided with limited idle time insertion, by exploiting a hybrid offline/online technique known as Precedence Constraint Posting (PCP). Our approach does not require probability distributions to be specified, relying instead on simple and cheaper-to-obtain information (bounds, average values). The method has been tested on synthetic applications/platforms and compared with an offline optimized Fixed Priority Scheduling (FPS) approach and a pure online FIFO scheduler; the results are very promising, as the PCP schedules exhibit good stability and improved average execution time (14 percent on average, up to 30 percent versus FPS and up to 40 percent versus the FIFO scheduler).

Optics of CdS/CdTe Thin-Film Photovoltaics

Optical reflection probes based on high-speed multichannel spectroscopic ellipsometry (SE) have been applied at different stages in the development and evaluation of CdS/CdTe thin-film photovoltaics (PV). Real-time SE during materials fabrication has provided insights into the nucleation, coalescence, and structural evolution of CdS and CdTe films. The deduced optical properties of the materials have provided insights into material density, stress, and excited carrier mean free path or defect density. The optical properties have also served as a database for analyzing complete PV stacks by probing through the glass. Uses of the database include CdTe materials analysis by Br <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> /methanol etching for depth profiling of devices before and after CdCl <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> treatment, as well as large-area PV plate and module analysis by off-line mapping with future on-line capabilities.

Estimation of cycle-resolved in-cylinder pressure and air-fuel ratio using spark plug ionization current sensing

In recent years, several new sensor technologies have been developed and implemented within automotive industries due to the increasing requirements for improved engine performance and emission reduction. It requires detailed and specified knowledge of the combustion process inside the engine cylinder along with a sophisticated technique in engine diagnostics and control. During the last few years, the ionization current signal detection has been the emerging technology in the new sensor developments, in which the spark plug is used as a combustion probe, to improve the performance and emissions of an automobile engine. In this paper, a novel methodology will be presented which allows the cycle-resolved as well as the mean-value estimation of the air-fuel ratio and in-cylinder pressure based on the ionization current signal measurements. The implementation details of this methodology as well as extensive results will be presented for a wide range of air-fuel ratios. The main advantage of this new approach to process the ionization signal is its strong potential for real-time estimation of the air-fuel ratio and combustion diagnostics of individual cylinders and engine cycles. All the complex physics during the actual events (combustion process, ion generation, engine dynamics, etc.) are automatically self-extracted by this technique from acquired data in an initial off-line mapping phase. Once this has been performed, the air-fuel ratio and in-cylinder pressure can easily be estimated for each individual cylinder and combustion event in real-time with few computational requirements. Hence, this methodology has a high potential for the real-time combustion diagnostics and engine control based on the air-fuel ratio and in-cylinder pressure, while eliminating the requirements for installing expensive air-fuel ratio and in-cylinder pressure sensors. The results indicate that estimation of the cycle-resolved air-fuel ratio and in-cylinder pressure is reasonably accurate and robust, despite the inherently noisy character of the ionization signals, with estimation errors typically in the order of 2 per cent or less, except for very fuel-rich conditions.

ALEXIS Spacecraft Attitude Reconstruction with Thermal/Flexible Motions Due to Launch Damage

An attitude reconstruction algorithm has been developed and tested for the ALEXIS spacecraft, whose dynamics include significant thermal/flexible motions due to structural damage that occurred during launch. The ALEXIS science mission needs the reconstructed attitude to do off-line mapping of soft x-ray photons from telescopes' image planes aboard the spinning spacecraft into inertial coordinates. The attitude reconstruction algorithm takes inputs from a fine sun sensor and a horizon crossing indicator and processes them using a square-root-information filter/smoother. The filter/ smoother model includes Euler dynamics and randomwalk torques to approximate thermal/flexible effects. The launch damage left a solar array paddle free to flap under the influence of thermal inputs, which causes spin rate and direction changes. Despite the complexity and uncertainty of the dynamic model, the algorithm achieves a 0.25° 1-cr accuracy over 30-min. periods of orbit night, when attitude data is unavailable.