Search

The increasing need for efficient and sustainable agriculture has driven the adoption of advanced technologies like Artificial Intelligence (AI) in crop management. This study evaluates the feasibility of deploying AI models on ARM-M microcontrollers for real-time detection of rice plant diseases, crucial for early intervention and crop health management. We developed and assessed the performance of five Convolutional Neural Network (CNN) architectures: MobileNet V1, MobileNet V2, FD-MobileNet, ResNet, and SqueezeNet. These models were trained on a dataset comprising 5932 high-quality images of rice leaves affected by four prevalent diseases: bacterial blight, blast, brown spot, and tungro. The AI models were optimized and quantized to meet the computational and memory constraints of the ARM-M microcontroller. Our evaluation focused on accuracy, computational efficiency, and resource utilization. The results demonstrated that FD-MobileNet achieves high accuracy of 98.44% with lower computational costs, making it suitable for deployment in resource-constrained environments. The deployment on ARM-M microcontrollers showcased the potential for real-time, on-site disease monitoring, enabling farmers to take timely and effective actions. This study underscores the transformative impact of integrating AI with embedded systems in smart agriculture, promoting sustainable farming practices and enhancing food security.

Foveated rendering (FR)—in which the central foveal layer (the area around the eye gaze) of a virtual reality (VR) image is rendered at the highest resolution and the peripheral layers are rendered at progressively lower resolutions—is an advanced VR technique that controls the balance between computational load and visual quality by adjusting the foveal layer sizes. We consider an edge computing-assisted VR computation offloading system incorporating FR and develop a deep reinforcement learning (DRL)-based solution to maximize a unified objective combining the visual quality and the overheads of energy consumption and delay for multiple VR users by optimizing the foveal layer size determination, offloading decisions and radio resource allocation of wireless links with non-orthogonal multiple access (NOMA). To formulate the unified objective, the user-perceived visual quality of a VR image rendered via FR is modeled with the sizes of the foveal layers as variables. The proposed solution consists of three main modules: per-user FR and offloading modules, which determine the sizes of the foveal layers and make offloading decisions, including subband allocation, and a central transmit-power allocation module. The actions of these three modules determine the reward that is in turn fed back to each module. Evaluation results reveal that the proposed framework can better adapt the operations of FR, offloading and wireless transmission to the environmental conditions than other FR and offloading benchmarks in terms of overall reward, visual quality, energy consumption and delay.

Offloading virtual reality (VR) computations to a cloud computing entity can enable support for VR services on low-end user devices but may result in increased latency, which will lead to mismatch between the user’s viewport and the received VR image, thus inducing motion sickness. Predicting future motion and rendering future images accordingly is a promising solution to the latency problem. In this paper, we develop velocity- and error-aware model switching schemes applicable to a wide range of existing motion prediction models. First, we consider the chattering problem of machine learning (ML)-based prediction models and the relationship between the velocity and the prediction error gap between an ML model and the case of no prediction (NOP). Accordingly, we propose a velocity-aware switching (VAS) scheme that combines the outputs from the ML model and the NOP case via a weight determined by the head motion velocity. Next, we develop an ensemble method combining a set of outputs from VAS and other models, called error-aware switching (EAS). EAS switches between model outputs based on the error statistics of those outputs under the parallel execution of multiple models, including VAS models. For EAS, schemes for both hard switching and soft integration of the model outputs are proposed. We evaluate the proposed schemes based on real VR motion traces for diverse ML-based prediction models.

Eye tracking is a technology that is in high demand, especially for next-generation virtual reality (VR), because it enables foveated rendering, which significantly reduces computational costs by rendering only the area at which a user is gazing at a high resolution and the rest at a lower resolution. However, the conventional eye-tracking technique requires per-eye camera hardware attached near the eyes within a VR headset. Moreover, the detected eye gaze follows the actual eye gaze with a finite delay because of the camera latency, the need for image processing, and the VR system’s native latency. This paper proposes an eye-tracking solution that predicts a user’s future eye gaze using only the inertial sensors that are already built into VR headsets for head tracking. To this end, we formulate three time-series regression problems to predict (1) the current eye gaze using past head orientation data, (2) the future eye gaze using past head orientation and eye gaze data, and (3) the future eye gaze using past head orientation data only. We solve the first and second problems using machine learning models and develop two solutions for the final problem: two-stage and single-stage approaches. The two-stage approach for the final problem relies on two machine learning models connected in series, one for the first problem and the other for the second problem. The single-stage approach uses a single model to predict the future eye gaze directly from past head orientation data. We evaluate the proposed solutions based on real eye-tracking traces captured from a VR headset for multiple test players, considering various combinations of machine learning models. The experimental results show that the proposed solutions for the final problem reduce the error for a center-fixed gaze by up to 50% and 20% for anticipation times of 50 and 150 ms, respectively.

Offloading heavy virtual reality (VR) computational operations to a network edge computation entity is receiving increasing attention as a tool to wirelessly and energy efficiently provide low-end client devices with high-quality and immersive interactive VR services anytime and anywhere across the globe. In this work, we aim to provide an understanding of various characteristics of VR computation offloading through comprehensive experiments conducted using a prototype testbed for edge-assisted VR processing and streaming. First, we investigate the benefits of VR offloading in terms of computational load and power consumption reduction for a client device compared to standalone operation. Next, we measure VR traffic patterns, including frame size and data and packet rates with various settings, such as different resolution and encoding options. We also measure several performance metrics associated with the quality of experience, namely, frame rate, packet loss rate, and image quality, with various configuration settings. Then, we present latency measurement studies and investigate per-component latency with various settings. Furthermore, we report the rigorous experiments performed to study the impacts of latency and motion patterns on the black borders formed due to image reprojection and the overfilling technique used to eliminate these black borders.

The wheelchair is one of the most common assistive technologies for people with motor impairment, due to its environmentally friendly features in terms of mobility and comfort. However, the operational method for the conventional wheelchair is still inconvenience for people with finger problems. Therefore, in the project, a novel control method to operate an electric wheelchair, using hand gestures (wrist rotation) is developed. Sixty-five (65) participants were involved in this study. Five hand gestures were considered and studied for forward, backward, right, left, and stop maneuvers while considering the human ergonomics factor. In this study, the stop maneuver was determined based on the most comfortable hand position to mitigate a fatigue experience, with two gesture classifying methods further investigated. The first method based on threshold has a promising accuracy of 96% and 91% precision. This method, however, requires a calibration every time a new user is introduced. The second method, a Naïve Bayes approach, was observed to solve the problem, as it has about 99% of both accuracy and precision. The evaluation of this method was then conducted with six participants that operated the wheelchair to follow the trajectories from start to end. The results showed that the participants comfortably controlled the developed wheelchair system to the goal without any collision. Results from experiments indicate that the proposed approach has high accuracy and the potential to solve the problem related to finger dependencies and hand fatigue.

With the exponential rise of the number of IoT devices, the amount of data being produced is massive. Thus, it is unfeasible to send all the raw data directly to the cloud for processing, especially for data that is high dimensional. Training deep learning models incrementally evolves the model over time and eliminates the need to statically training the models with all the data. However, the integration of class incremental learning and the Internet of Things (IoT) is a new concept and is not yet mature. In the context of IoT and deep learning, the transmission cost of data in the edge-cloud architecture is a challenge. We demonstrate a novel sample selection method that discards certain training images on the IoT edge device that reduces transmission cost and still maintains class incremental learning performance. It can be unfeasible to transmit all parameters of a trained model back to the IoT edge device. Therefore, we propose an algorithm to find only the useful parameters of a trained model in an efficient way to reduce the transmission cost from the cloud to the edge devices. Results show that our proposed methods can effectively perform class-incremental learning in an edge-cloud setting.