United Technologies Research Center

Unmanned Aerial Vehicles (UAVs) are increasingly deployed within reconnaissance areas to collect sensor data. This tremendous amount of data is often uploaded to a central base station for analyzing and identifying potential targets. However, centralized approaches to Automatic Target Recognition (ATR) introduce problems such as network communication delays and susceptibility of the system to malicious attacks, thereby limiting autonomous capabilities. To overcome these challenges, we have developed an approach for artificial intelligence on the edge to enable reliable affordable mass. Specifically, we have developed a Continuous Deployment for Machine Learning (CD4ML) framework for operationalizing ML approaches, incorporating capabilities for inferencing, drift detection, model adaptation, and performance verification. We have deployed this framework on a real UAV to enable vision-based feedback control for ATR applications. The autonomy of the UAV is demonstrated in our laboratory for detecting, identifying, and tracking a ground vehicle. Finally, we have also developed a semi-automated, reusable, and modular workflow for the deployment of AI at the edge on target FPGA environments, as well as certification tools that will aid in further maturation of our developed approach for ATR applications.

In this paper we consider aspects of geometric observability for hypergraphs, extending our earlier work from the uniform to the nonuniform case. Hypergraphs, a generalization of graphs, allow hyperedges to connect multiple nodes and unambiguously represent multi-way relationships which are ubiquitous in many real-world networks including those that arise in biology. We consider polynomial dynamical systems with linear outputs defined according to hypergraph structure, and we propose methods to evaluate local, weak observability.

Insider threats are considered a major threat to information and communication technology (ICT) systems creating an important source of vulnerabilities from a security perspective. The technical knowledge that insiders have about the ICT systems, such as its IT infrastructure, the high load of data generated by other employees of the company which hides insiders' activities, their access rights as well as the confidentiality of the data of which they have access to, creates the perfect scenario for a powerful yet undetected attack. State of the art techniques and security operations center tools struggle to come up with effective solutions to recognise these threats. Therefore, in this paper, we propose a novel artificial intelligence based constraint learning technique to help their detection. The approach creates an optimized constraint network representing the nominal behaviour of an employee and detects threatening events when their associated costs are above a certain threshold. The threshold is learnt alongside with the constraint network model. The proposed approach is based on detection models able to provide human interpretable feedback regarding the detection performed. These information are crucial in helping system operators to understand why the detection has occurred and to help them acting promptly on the threat. The explanation comes directly from the structure of the detection model and relies on the identification of which constraints are being violated. The approach is tested on the CERT insider threat dataset v4.2 and the results obtained look promising, achieving at least the same accuracy as other state of the art techniques as well as providing the details regarding the broken constraints of the threat. A comparison with state of the art techniques applied on this dataset is also provided, showing the strength of our results.

Traditionally safety in industrial manufacturing environments has been addressed by rigid separation between robots and human operators. With the advent of new technologies and the transition of production to industry 4.0, a more flexible approach to manufacturing is pursued to achieve higher productivity, where robots and human operators are allowed to collaborate and interact. This transformation leads to overcoming traditional safety procedures and the development of new safety-assuring technologies for the minimization of risks connected with human robot collaboration. In this work we focus on the prediction of movements of operators’ upper torso and arms by developing a method which combines data driven methodologies with formal methods. The approach is based on a predictive model of human motion compared against the planned robot trajectory and online monitoring of satisfaction of safety requirements with formal methods. We provide an early assessment of the robustness of the proposed method with results from simulations in a virtual environment.

This papera describes a maintenance scheduling application, which was developed together with an industrial partner. This is a highly combinatorial decision process, to plan and schedule the work of a group of travelling repair technicians, which perform preventive and corrective maintenance tasks at customer locations. Customers are located both in urban areas, where many customers are in close proximity, and in sparsely populated rural areas, where the travel time between customer sites is significant. To balance the workload for the agents, we must consider both the productive working time, as well as the travel between locations. As the monolithic problem formulation is unmanageable, we introduce a problem decomposition into multiple sequential steps, that is compatible with current management practice. We present and compare different models for the solution steps, and discuss results on datasets provided by the industrial partner.

Wireless Sensor Network (WSN) technology has developed substantially over the past decade or so and now numerous solutions exist across a diverse range of innovative applications. The expanding Internet of Things (IoT) sector is becoming an ever more important aspect of modern technology and a key motivator for improving security and privacy in WSNs. Typically, WSN protocols form an integral part of the overall IoT infrastructure by enabling the sensor to access point communication links. These wireless links inherently encompass security challenges, frequently due to external interference and intrusions. As IoT applications incorporate WSNs in their architecture, the incentive to attack and compromise these WSNs escalates. Often, commercial off the shelf devices and standardized open-access protocols combine to achieve specific WSN deployments. Numerous WSN vulnerabilities exist, whilst attack approaches are abundant and change frequently. Thus, to ensure acceptable performance, safety and privacy in many IoT applications, the adopted WSN must be secure. This paper discusses IoT security and privacy, by evaluating a machine learning approach for interference detection focused entirely on analyzing received In-phase (I) and Quadrature-phase (Q) samples. Significantly, once an intrusion is detected, mitigation strategies can be implemented, thus emphasizing the requirement for interference detection. Random Forest is chosen as the machine learning classifier as it consists of a large number of individual decision trees operating as an ensemble. An intrusion detection system (IDS) is developed based on Matlab simulated ZigBee data as an initial insight into whether a real wireless data approach may be viable.

Glioblastoma multiforme (GBM) is a fast-growing and highly invasive brain tumor, which tends to occur in adults between the ages of 45 and 70 and it accounts for 52 percent of all primary brain tumors. Usually, GBMs are detected by magnetic resonance images (MRI). Among MRI, a fluid-attenuated inversion recovery (FLAIR) sequence produces high quality digital tumor representation. Fast computer-aided detection and segmentation techniques are needed for overcoming subjective medical doctors (MDs) judgment. This study has three main novelties for demonstrating the role of topological features as new set of radiomics features which can be used as pillars of a personalized diagnostic systems of GBM analysis from FLAIR. For the first time topological data analysis is used for analyzing GBM from three complementary perspectives—tumor growth at cell level, temporal evolution of GBM in follow-up period and eventually GBM detection. The second novelty is represented by the definition of a new Shannon-like topological entropy, the so-called Generator Entropy. The third novelty is the combination of topological and textural features for training automatic interpretable machine learning. These novelties are demonstrated by three numerical experiments. Topological Data Analysis of a simplified 2D tumor growth mathematical model had allowed to understand the bio-chemical conditions that facilitate tumor growth—the higher the concentration of chemical nutrients the more virulent the process. Topological data analysis was used for evaluating GBM temporal progression on FLAIR recorded within 90 days following treatment completion and at progression. The experiment had confirmed that persistent entropy is a viable statistics for monitoring GBM evolution during the follow-up period. In the third experiment we developed a novel methodology based on topological and textural features and automatic interpretable machine learning for automatic GBM classification on FLAIR. The algorithm reached a classification accuracy up to 97%.

Liquid electrolyte solutions containing dissolved active material imbue Redox Flow Batteries (RFBs) with many unique benefits relative to solid-state electrochemical energy storage devices: long shelf-life, long system life-time, de-coupled rated energy and power, ease of recyclability.1 Operating a RFB system with flowing electrolyte solutions presents unique concerns as well, with one notable concern arising from possible precipitation of electrolyte. Precipitation of active material can have many consequences including sensor inaccuracy, inaccessible storage capacity, partially- or fully-blocked reactor electrode regions, and system clogging.Aqueous vanadium RFBs, a leading chemistry in commercial systems, has a precipitation mechanism at high temperatures for battery solutions containing commercially relevant vanadium concentrations >1.5 M. At elevated temperatures (> 40°C), positive electrolyte solutions can precipitate V2O5 dimers, first as a gel-like substance and eventually as solid particles.2 The UTRC team has developed recovery procedures for systems that may succumb to such precipitation events; in the case for aqueous vanadium RFBs, exposure of precipitate to lower oxidation state forms of vanadium (i.e. negative electrolyte) reduces and redissolves precipitated V2O5. This electrolyte takeover process (ETP) can successfully redissolve precipitates and recover performance from precipitate-induced cell/stack under-performance.As UTRC pursues alternative flow chemistries, the same ETP principles are applied to other chemistries such as polysulfide-based systems, known for sulfur crossover and precipitation of elemental sulfur in positive electrode volumes.3 This abstract will discuss the progress made at UTRC to recover from precipitation events in such chemistries, which, unlike aqueous vanadium, do not permit simple mixing of negative and positive electrolytes. Acknowledgements: The authors gratefully acknowledge support from Vionx Energy Corporation. Some of the work to be presented here was funded by the Advanced Research Projects Agency - Energy (ARPA-E), U.S. Department of Energy (DOE) under Award Number DE-AR0000994.

Given a Boolean formula ϕ(x) in conjunctive normal form (CNF), the density of states counts the number of variable assignments that violate exactly e clauses, for all values of e. Thus, the density of states is a histogram of the number of unsatisfied clauses over all possible assignments. This computation generalizes both maximum-satisfiability (MAX-SAT) and model counting problems and not only provides insight into the entire solution space, but also yields a measure for the hardness of the problem instance. Consequently, in real-world scenarios, this problem is typically infeasible even when using state-of-the-art algorithms. While finding an exact answer to this problem is a computationally intensive task, we propose a novel approach for estimating density of states based on the concentration of measure inequalities. The methodology results in a quadratic unconstrained binary optimization (QUBO), which is particularly amenable to quantum annealing-based solutions. We present the overall approach and compare results from the D-Wave quantum annealer against the best-known classical algorithms such as the Hamze-de Freitas-Selby (HFS) algorithm and satisfiability modulo theory (SMT) solvers.

Al deposition from chloroaluminate ionic liquids necessitates the use of consumable anodes and near 100% plating efficiency in order prevent irreversible breakdown of the electrolyte. A significant side effect of this high plating efficiency is poor deposit thickness uniformity across current density gradients. For simple or axisymmetric geometries under ideal plating conditions it is straight forward to mitigate such issues through traditional thieving or shielding. However, as part geometries become more complex, and plating conditions less ideal (i.e. barrel plating of small parts), methods of improving poor throwing power resulting from secondary current density effects become less obvious. Addressing this issue is critical in enabling Al plating technologies to progress commercially. We have explored the effects of metal concentration, additives and plating waveform on deposit thickness uniformity for rack and barrel plated parts of increasing geometric complexity in order to minimize plating thickness distribution and improve throwing power of [AlCl3:EMIMCl] chloroaluminate electrolytes into high aspect ratio features.