In space exploration and communication technology, software-defined networking architectures significantly improve the communications throughput and latency. This study developed a cognitive agent based on Q-learning with continuous learning capabilities to select optimal routing options in a dynamically changing networking environment. This proposed method achieves up to 90% reduction in measured Internet Control Message Protocol (ICMP) round-trip time compared to the Dijkstra shortest-path baseline under identical test conditions. Under dynamically changing network link latency conditions, the cognitive agent achieved a 90% success rate in routing packets in laboratory experiments. The agent was executed on neuromorphic hardware called Intel Loihi. However, due to the system’s power limitations, a simplified version of the agent was engineered to launch into space aboard a CubeSat. The CubeSat was launched in January 2022, making a historic milestone as the first launch of a neuromorphic system into space. The developed applications were successfully executed in space.

To effectively transmit data to and from satellites requires a complex and robust RF communication system. Using adaptive link techniques, several different types of signal modulations may be used to maximize satellite efficiency depending on a variety of unexpected channel impairments. We propose an FPGA based neural network system capable of learning these RF signal modulations using a supervised learning technique designed for low power, high-efficiency in-orbit deployment. Thus, as signal modulation changes, a receiving satellite would be able to recognize this modulation update without adjustment from an operator, or any delay due to control signal transmission.Due to the limited resources of FPGAs, we implement CNN transfer learning on the FPGA instead of full CNN training. The work presented demonstrates a convolutional neural network (CNN) capable of learning and recognizing a set of modulation schemes commonly used to transmit RF information. We are capable of recognizing the modulation scheme from the I and Q data channels directly, with no preprocessing or data conversion required other than segmenting the incoming signal into a set of uniform normalized samples. We perform a network design and size analysis, showing that we can obtain reasonably high accuracy using networks with a relatively low number of trainable parameters. Thus, modulation recognition can be performed through rapid feedforward computation. Given that a user of this system may wish to receive a signal using a modulation scheme that the network has not previously learned, we demonstrate that transfer learning can learn new modulation schemes by retraining only the fully connected layers in the CNN. Thus, this type of network would excel in outer space deployment using high-efficiency transfer learning hardware. Our FPGA system utilized 2.1W power and processed 219 data samples per second for training. This made it about 59x more energy efficient than a desktop GPU for the same task.

Network intrusion detection is crucial for securing data transmission against cyber threats. Traditional anomaly detection systems use computationally intensive models, with CPUs and GPUs consuming excessive power during training and testing. Such systems are impractical for battery-operated devices and IoT sensors, which require low-power solutions. As energy efficiency becomes a key concern, analyzing network intrusion datasets on low-power hardware is vital. This paper implements a low-power anomaly detection system on Intel's Loihi and Brainchip's Akida neuromorphic processors. The model was trained on a CPU, with weights deployed on the processors. Three experiments-binary classification, attack class classification, and attack type classification-are conducted. We achieved approximately $\mathbf{9 8. 1 \%}$ accuracy on Akida and $\mathbf{9 4 \%}$ on Loihi in all experiments while consuming just 3 to 6 microjoules per inference. Also, a comparative analysis with the Raspberry Pi 3 and Asus Tinker Board is performed. To the best of our knowledge, this is the first performance analysis of low power anomaly detection based on spiking neural network hardware.

We introduce VelocitySNN-Fuzzy AI, a novel neuromorphic system for real-time determination of three-dimensional velocity vectors (Vx, Vy, Vz) during Martian drone flight. Unlike Earth-Based UAVs, Mars operations cannot rely on GPS or conventional visual odometry due to the lack of infrastructure and the extreme lighting and terrain conditions. Our approach integrates a bio-inspired event-based camera, a Spiking Neural Network (SNN) developed using SpikingJelly, and an explainable Fuzzy AI Logic layer to produce reliable and interpretable velocity determination in challenging Mars-Like environments. Validated using publicly available Mars-Style drone datasets, the architecture demonstrates low-latency, high-dynamic-range perception suitable for small, power-constrained UAV platforms. VelocitySNN-Fuzzy AI determines 3D velocity for Martian drones using event-based vision, spiking neural networks, and fuzzy logic. Designed for GPS-denied, low-light environments, it enables real-time, low-power navigation. Validated on FPV drone datasets, it supports autonomous flight with interpretable outputs and deployment on neuromorphic hardware like Loihi and Akida.

In this work we present neural network algorithms for learning several marketing and business analytics datasets. We study four different datasets for potential applications in the business analytics realm, which apply to marketing, customer screening, customer churn, and credit card security. For each of these datasets we examine which neural network topology is likely to provide the best model for prediction. We also examine data preprocessing techniques to determine which features in the datasets are most likely contributing to successful prediction, and we also address class imbalance for datasets where data entries of interest occur infrequently in a dataset. Our results show training and testing accuracies between 80% and 100% depending on the dataset examined. We use these experiments to demonstrate how implementing these machine learning methods can reduce business operating costs for those adopting this technology. Thus, even in the cases of lower accuracy (about 80%) we can still justify significant benefit to the user based on time saved using these neural networks as pre-screening tools within their workflow.

An ultralow-power, high-performance online-learning and anomaly-detection system has been developed for edge security applications. Designed to support personalized learning without relying on cloud data processing, the system employs sample-wise learning, eliminating the need for storing entire datasets for training. Built using memristor-based analog neuromorphic and in-memory computing techniques, the system integrates two unsupervised autoencoder neural networks—one utilizing optimized crossbar weights and the other performing real-time learning to detect novel intrusions. Threshold optimization and anomaly detection are achieved through a fully analog Euclidean Distance (ED) computation circuit, eliminating the need for floating-point processing units. The system demonstrates 87% anomaly-detection accuracy; achieves a performance of 16.1 GOPS—774× faster than the ASUS Tinker Board edge processor; and delivers an energy efficiency of 783 GOPS/W, consuming only 20.5 mW during anomaly detection.

Cognitive agents make systems autonomous through the process of decision automation by mining an existing knowledge repository at run time. These processes can often be highly compute intensive, and would thus run slowly on the low-power computing platforms typically seen in autonomous systems. This paper examines how knowledge be represented in a Q-table and proposes a novel fast algorithm to mine that knowledge based on constraints. We evaluate this approach for the knowledge mining process of a specific agent: Cognitively Enhanced Complex Event Processing (CECEP). Within CECEP, knowledge is represented using Cognitive Domain Ontologies (CDO), and is mined using situational inputs and constraints. This is a novel approach to store information and is able to accommodate CDOs with millions of solutions. To show that the approach can run on low power hardware in real-time, this algorithm was executed on two low-power minicomputing platforms - Intel’s NUC and Asus’s Tinker Board. At present, no other optimized CDO solvers can generate solutions on these platforms. The algorithm generated the same amount of solutions as a GPU-enabled optimized path-based forward checking CDO solver, while consuming around 7.7 and 5.15 times less energy (Joules) on the NUC and Tinker Board respectively.

The unprecedented progress in artificial intelligence (AI), particularly in deep learning algorithms with ubiquitous internet connected smart devices, has created a high demand for AI computing on the edge devices. This review studied commercially available edge processors, and the processors that are still in industrial research stages. We categorized state-of-the-art edge processors based on the underlying architecture, such as dataflow, neuromorphic, and processing in-memory (PIM) architecture. The processors are analyzed based on their performance, chip area, energy efficiency, and application domains. The supported programming frameworks, model compression, data precision, and the CMOS fabrication process technology are discussed. Currently, most commercial edge processors utilize dataflow architectures. However, emerging non-von Neumann computing architectures have attracted the attention of the industry in recent years. Neuromorphic processors are highly efficient for performing computation with fewer synaptic operations, and several neuromorphic processors offer online training for secured and personalized AI applications. This review found that the PIM processors show significant energy efficiency and consume less power compared to dataflow and neuromorphic processors. A future direction of the industry could be to implement state-of-the-art deep learning algorithms in emerging non-von Neumann computing paradigms for low-power computing on edge devices.

Spiking neural network (SNN) hardware has gained significant interest due to its ability to process complex data in size, weight, and power (SWaP) constrained environments. Memristors, in particular, offer the potential to enhance SNN algorithms by providing analog domain acceleration with exceptional energy and throughput efficiency. Among the current SNN architectures, the Liquid State Machine (LSM), a form of Reservoir Computing (RC), stands out due to its low resource utilization and straightforward training process. In this paper, we present a custom memristor-based LSM circuit design with an online learning methodology. The proposed circuit implementing the LSM is designed using SPICE to ensure precise device level accuracy. Furthermore, we explore liquid connectivity tuning to facilitate a real-time and efficient design process. To assess the performance of our system, we evaluate it on multiple datasets, including MNIST, TI-46 spoken digits, acoustic drone recordings, and musical MIDI files. Our results demonstrate comparable accuracy while achieving significant power and energy savings when compared to existing LSM accelerators. Moreover, our design exhibits resilience in the presence of noise and neuron misfires. These findings highlight the potential of a memristor based LSM architecture to rival purely CMOS-based LSM implementations, offering robust and energy-efficient neuromorphic computing capabilities with memristive SNNs.

Detecting anomalies and faults swiftly and accurately is essential in many applications, such as healthcare, infrastructure, industry, and security. In most of these applications, the available power for running an anomaly detection system is very limited. This is especially the case for modern smart and electric vehicles, which have an increasing amount of electronics. These vehicles have Electronic Control Units (ECUs) that communicate with each other through the Controller Area Network (CAN) bus. The Controller Area Network, however, is not very secure, so bad actors can do severe damage, putting the lives of drivers and passengers at risk. Thus, low-power anomaly detection is essential for both security and reliability. Deep learning networks have demonstrated proficiency in anomaly identification. However, their implementation on conventional Central Processing Units (CPUs) and Graphics Processing Units (GPUs) incurs significant energy consumption. To overcome this limitation, we propose a low-power approach for placing security directly in the network hardware. An autoencoder-based real-time anomaly detector trained through unsupervised learning is seamlessly mapped onto the Intel Loihi spiking neuromorphic processor. Our proposed anomaly detector used significantly less power than a CPU or GPU and showed the best accuracy and F-1 score to detect anomalies compared to the alternative spiking approaches. The proposed anomaly detection system consumed about 20,000 and 700 times less energy than a CPU and GPU respectively. To the best of our knowledge, this is the first low-power, unsupervised anomaly detection system using the Loihi or any other neuromorphic processor. The results of this work will apply to other anomaly detection applications, as well as other spiking neuromorphic processors, resulting in a universally applicable extreme low-power anomaly detection platform.

The event camera is researched, developed, and designed to imitate the human eye; it is a groundbreaking vision sensor with the following advantages over a standard camera: a net rate that is much faster, a latency that is far less, a high dynamic range, and it uses far less power. These fundamental properties assist in enabling the design of third-generation algorithms in Spiking Neural Networks intended to mimic the human brain and vision processing. Moreover, these fundamental properties enable swift robotics despite the challenges of motion blur and high latency that standard cameras face. Also, these properties should enable motion estimation from the surface features on Mars, which is difficult to achieve with standard cameras. For this research, the team is using a NASA Space use-case application. For this NASA Space application, the team has chosen a challenging motion-estimation task involving a Mars-based above-ground Helicopter beyond “Ingenuity” and the planet and surface of Mars. Event-based cameras have been gaining interest within the computer vision community. They are particularly suitable for applications with challenging temporal constraints and safety requirements. Thus, Event-based sensors are an excellent match for Spiking Neural Networks (SNNs), as coupling an asynchronous sensor with neuromorphic hardware can result in real-time systems with minimal power requirements. Moreover, methods to verify and validate event-based sensing platforms for space applications are lacking. In this paper, we investigate the addition of fuzzy logic models to arrive at an explainable SNN algorithm for the NASA space use-case. Ultimately, the team aims to develop a unified model yielding reasonably accurate optical flow estimates.

A memristor based neuromorphic processor for on-chip training is presented. Additionally, a novel approach utilizing in-situ learning to improve wireless signal modulation classification under adversarial jamming is described. The neuromorphic system is over 50× energy efficient than optimized digital systems at this wireless signal modulation task for similar accuracy levels.

Neuromorphic and deep learning (DL) algorithms are important research areas gaining significant traction of late. Due to this growing interest and the high demand for low-power and high-performance designs for running these algorithms, various circuits and devices are being designed and investigated to realize efficient neuromorphic and DL architectures. One device said to drastically improve this architecture is the memristor. In this chapter, studies investigating memristive implementations into neuromorphic and DL designs are summarized and categorized based on the switching mechanicsms of a few prominent memristive device technologies. Furthermore, the simulation platforms used to model both neuromorphic and DL hardware implementations, which use memristors, are summarized and discussed. This chapter can provide a quick reference for readers interested in learning the latest advancements in the areas of memristive devices and systems for use in neuromorphic and DL systems.

Software-defined networking (SDN) architectures can significantly improve communications throughput and latency. This study developed a Q-learning based cognitive agent on the Intel Loihi for SDN applications. We utilize continuous learning to enable optimal route selection in continuously changing networking environments. The Loihi based agent was able to route packets to the most optimal path 90% of the time with continuously changing network link latency. Due to system power constraints, a simplified version of the agent was then developed for launch into space aboard a CubeSat. The CubeSat was launched in January 2022 and the applications developed ran successfully in space. This is the first launch of a neuromorphic system performing SDN operation into space.

The ultra-wideband, frequency agile and high duty cycle features of Low Probability of Intercept (LPI) radar pose serious concerns for passive radar detection systems that must efficiently identify LPI radar modulation schemes with very low latency and high accuracy. For these tasks, conventional deep learning techniques have often been employed, though this comes at the cost of increased model complexity and high energy consumption when implemented in tactically relevant scenarios with significant channel deterioration and size, weight, and power (SWaP) constrained platforms. In this paper, we propose a neuromorphic approach based on energy-efficient and robust Spiking Neural Networks (SNNs). To demonstrate the effectiveness of our proposed approach, we perform a challenging LPI radar waveform recognition task with a Spiking Convolutional Neural Network (SCNN). Simulation results show that the SCNN achieves a 13-waveform recognition accuracy of 99.1% at OdB SNR with 1.51µJ of total energy per inference. It is found that our system outperforms non-spiking models with a reduction in model size and three orders of magnitude lower energy consumption. To the best of our knowledge, this work presents the first spike-based implementation of an LPI radar waveform recognition system. Moreover, this paper highlights the potential for extremely low SWaP neuromorphic hardware deployment to enable intelligent electronic support systems that demand high throughput efficiency and ultra-low power.

The use of wireless communication has been growing significantly. Given the limited radio spectrum available, it is important to use modulation techniques to squeeze the most information in the least amount of spectrum. To improve the throughput and reliability of communication, Automatic Modulation Classification (AMC) has become an essential topic in the research of Cognitive Radio (CR). AMC is a technique to identify the modulation scheme of the received radio signals without any transmission parameters. The state-of-the-art approach is to use a pre-trained classifier to determine the modulation in use. Unfortunately, all forms of distortions and interference are impossible to predict, and thus these approaches fail when new distortions or interference appear that were not previously learned. In this work, we developed a technique where the transmitter periodically sends out a known ordered sequence of signals of all the modulations to be used by the receiver. The receiver knows the transmission sequence and thus knows how to label these received signals containing each required modulation. We used an over-the-air (OTA) radio transmission between software-defined radio (SDR) devices, and the collected dictionary dataset was used to train a neural network at the receiver. Our results show that training the network in real-time while the radio is in use can enable significantly better performance than any of the existing approaches.

Unmanned aerial vehicles (UAVs), also known as drones, are expected to play an integral role in next-generation wireless avionics networks. To help secure these networks from malicious activity, machine learning (ML) based approaches have been proposed to identify drones and their flight modes via direct radio frequency signals. These approaches frequently rely on cloudcentric methods for training, which present serious concerns, such as privacy leakage, resource burden, and undesirable latency. In response to these concerns, a distributed ML paradigm known as federated learning (FL) has been proposed that enables multiple drones to collaboratively train ML models by only exchanging model parameters and not the raw data itself. Unfortunately, the conventional FL framework is strongly dependent on a centralized aggregation server with multiple resource-constrained edge devices, making it susceptible to poisoning attacks and poor network utilization. Recently, it has been found that blockchain technology (BT) holds significant promise for securing and storing data in edge applications with high levels of trust. However, the expanding chain of security blocks consumes a significant amount of computing power, thus limiting its scalability. To bridge the gap between computational efficiency and security, we propose a blockchain-empowered and energy-efficient FL framework with neuromorphic edge devices. To assess the effectiveness of our proposed framework we perform a drone identification and flight mode detection task with a spiking neural network (SNN). Finally, we compare our neuromorphic approach with competitive alternatives to validate the energy and performance gains of our system.

Reinforcement learning (RL) has shown its viability to learn when an agent interacts continually with the environment to optimize a policy. This work presents a memristor-based deep reinforcement learning (Mem-DRL) system for on-chip training, where the learning process takes place in a dynamic cartpole environment. Memristor device variability is taken into account to make the study more realistic. The proposed system utilized an analog ReLU module to reduce analog to digital converter usage. The analog Mem-DRL system consumed 191 times less energy than an optimized digital FP16 computing system. Our Mem-DRL system reduced the ADC usages by 40%, which led to reduced the overall system energy by 42%. Mem-DRL is 2.4 times faster than the FP16 system and performs 9.27 GOPS during DRL training. The system exhibited an energy efficiency of 23.8 TOPS/W.

Spiking neural networks represent a transition from deep networks and tensor engines to more dynamic systems better suited for carrying out decisions based on temporal patterns within data. Furthermore, memristor based hardware for synaptic computation provides the potential for dramatic gains in terms of portability, power reduction, and throughput efficiency. In this paper we combine these ideas and present a memristor crossbar based implementation of a liquid state machine based on spiking neurons. The design is verified by demonstrating successful classification of MIDI (Musical Instrument Digital Interface) signals. Given the temporal nature of musical works, this dataset is well suited to show the effectiveness of the proposed hardware. The proposed circuit implementing this LSM is designed using SPICE to ensure accuracy at the device level, which aids in detailed analysis and circuit optimizations. Liquid layer activity is converted to state vectors using custom hardware that does not bottleneck throughput. Along with the proposed architecture, a neuron modeling procedure is described to match the software and hardware spiking neuron outputs, proving the feasibility of a full-scale implementation of the LSM system. Furthermore, various non-idealities are analyzed in the network, including misfiring neurons and noise associated with updating memristor conductance. We find that the proposed architecture is resilient to noise and fault-tolerant. In simulation, the performance of the memristor-based LSM is demonstrated on a musical genre classification experiment, where a recognition accuracy of 90% is achieved, compared to 91% accuracy in software. To the best of our knowledge this work represents the first end to end circuit implementation of an analog spiking neural network based reservoir computing system where memristors are responsible for all computation, with custom CMOS hardware for sampling the liquid layer. Finally, we tie the work together with an energy and timing analysis of the proposed system.

Artificial Intelligence (AI) is moving towards the edge. Training an AI model for edge computing on a centralized server increases latency, and the privacy of edge users is jeopardized due to private data transfer through a less secure communication channels. Additionally, existing high-power computing systems are battling with memory and data transfer bottlenecks between the processor and memory. Federated Learning (FL) is a collaborative AI learning paradigm for distributed local devices that operates without transferring local data. Local participant devices share the updated network parameters with the central server instead of sending the original data. The central server updates the global AI model and deploys the model to the local clients. As the local data resides only on the edge, these devices need to be protected from cyberattacks. The Federated Intrusion Detection System (FIDS) could be a viable system to protect edge devices as opposed to a centralized protection system. However, on-device training of the model in resource constrained devices may suffer from excessive power drain, in addition to memory and area overhead. In this work we present a memristor based system for AI training on edge devices. Memristor devices are ideal candidates for processing in memory, as their dynamic resistance properties allow them to perform multiply-add operations in parallel in the analog domain with extreme efficiency. Alternatively, existing CMOS-based PIM systems are typically developed for edge inference based on pretrained weights, and are not equipped for on-chip training. We show the effectiveness of the system, where successful learning and recognition is achieved completely within edge devices. The classification accuracy of the memristor system shows negligible loss when compared a software implementation. To the best of our knowledge, this first demonstration of a memristor based federated learning system. We demonstrate the effectiveness of this system as an intrusion detection platform for edge devices, although given the flexibility of the learning algorithm, it could be used to enhance many types of on board leaning and classification applications.