A Joint Model Provisioning and Request Dispatch Solution for Low-Latency Inference Services on Edge

With the advancement of machine learning, a growing number of mobile users rely on machine learning inference for making time-sensitive and safety-critical decisions. Therefore, the demand for high-quality and low-latency inference services at the network edge has become the key to modern intelligent society. This paper proposes a novel solution that jointly provisions machine learning models and dispatches inference requests to reduce inference latency on edge nodes. Existing solutions either direct inference requests to the nearest edge node to save network latency or balance edge nodes’ workload by reducing queuing and computing time. The proposed solution provisions each edge node with the optimal number and type of inference instances under a holistic consideration of networking, computing, and memory resources. Mobile users can thus be directed to utilize inference services on the edge nodes that offer minimal serving latency. The proposed solution has been implemented using TensorFlow Serving and Kubernetes on an edge cluster. Through simulation and testbed experiments under various system settings, the evaluation results showed that the joint strategy could consistently achieve lower latency than simply searching for the best edge node to serve inference requests.

IMRP: interference-aware multicasting for multiple flows in WMNs.

Multicast is an important routing service for many of fundamental applications such as Webcasting, Video Streaming or Conferencing, Online Games. On the other hand, as much as Wireless Mesh Networks (WMNs) are growing and become more complex, the need for modification of Multicast Routing for WMNs with different constraints are increased as well. Therefor Researches on multicasting in WMN is still in initial stage. We focused on Multicasting over Single Channel WMNs with multiple group flows scenario. The main challenge on this scenario is the impact of interference between multiple flows on the performance of multicast flows. By increasing number of receiver nodes in a single channel area, the interference caused by multiple simultaneous flows in the network cannot be fully avoidable, but implementing a proper routing helps to improve the performance of the flows. In this research, we develop a new interference-aware routing metrics and implement it in our purposed routing scheme (IMRP) for multicasting multiple flows in WMNs. Our purposed metrics takes some new defined attributes of routing path into account. The purposed scheme is evaluated by simulating on Riverbed Modeller 17.5. Simulation and testbed experiments show that our proposed routing significantly improves the total Throughput as well as total Packet Delivery Ratio of multicast flows compared to existing standards.

IMRP: interference-aware multicasting for multiple flows in WMNs.

Multicast is an important routing service for many of fundamental applications such as Webcasting, Video Streaming or Conferencing, Online Games. On the other hand, as much as Wireless Mesh Networks (WMNs) are growing and become more complex, the need for modification of Multicast Routing for WMNs with different constraints are increased as well. Therefor Researches on multicasting in WMN is still in initial stage. We focused on Multicasting over Single Channel WMNs with multiple group flows scenario. The main challenge on this scenario is the impact of interference between multiple flows on the performance of multicast flows. By increasing number of receiver nodes in a single channel area, the interference caused by multiple simultaneous flows in the network cannot be fully avoidable, but implementing a proper routing helps to improve the performance of the flows. In this research, we develop a new interference-aware routing metrics and implement it in our purposed routing scheme (IMRP) for multicasting multiple flows in WMNs. Our purposed metrics takes some new defined attributes of routing path into account. The purposed scheme is evaluated by simulating on Riverbed Modeller 17.5. Simulation and testbed experiments show that our proposed routing significantly improves the total Throughput as well as total Packet Delivery Ratio of multicast flows compared to existing standards.

Multi-Filter Decoding in WiFi Backscatter Communication

WiFi backscatter communication has emerged as a promising enabler of ultralow-power connectivity for Internet of things, wireless sensor network and smart energy. In this paper, we propose a multi-filter design for effective decoding of WiFi backscattered signals. Backscattered signals are relatively weak compared to carrier WiFi signals and therefore require algorithms that filter out original WiFi signals without affecting the backscattered signals. Two multi-filter designs for WiFi backscatter decoding are presented: the summation and delimiter approaches. Both implementations employ the use of additional filters with different window sizes to efficiently cut off undesired noise/interference, thus enhancing frame detection and decoding performance, and can be coupled with a wide range of decoding algorithms. The designs are particularly productive in the frequency-shift WiFi backscatter communication. We demonstrate via prototyping and testbed experiments that the proposed design enhances the performance of various decoding algorithms in real environments.

Motion-Aware Interplay between WiGig and WiFi for Wireless Virtual Reality

Wireless virtual reality (VR) is a promising direction for future VR systems that offloads heavy computation to a remote processing entity and wirelessly receives high-quality streams. WiGig and WiFi are representative solutions to implement wireless VR; however, they differ in communication bandwidth and reliability. Our testbed experiments show that the performance of WiGig and VR traffic generation strongly correlates with and consequently can be predicted from a user’s motion. Based on this observation, we develop a wireless VR system that exploits the benefits of both links by switching between them and controlling the VR frame encoding for latency regulation and image quality enhancement. The proposed system predicts the performance of the links and selects the one with a higher capacity in an opportunistic manner. It adjusts the encoding rate of the host based on the motion-aware prediction of the frame size and estimated latency of the selected link. By evaluating the testbed data, we demonstrate that the proposed system outperforms a WiGig-only system with a fixed encoding rate in terms of latency regulation and image quality.

The 23 June 2016 West Virginia Flash Flood Event as Observed through Two Hydrometeorology Testbed Experiments

The flash flood event of 23 June 2016 devastated portions of West Virginia and west-central Virginia, resulting in 23 fatalities and 5 new record river crests. The flash flooding was part of a multiday event that was classified as a billion-dollar disaster. The 23 June 2016 event occurred during real-time operations by two Hydrometeorology Testbed (HMT) experiments. The Flash Flood and Intense Rainfall (FFaIR) experiment focused on the 6–24-h forecast through the utilization of experimental high-resolution deterministic and ensemble numerical weather prediction and hydrologic model guidance. The HMT Multi-Radar Multi-Sensor Hydro (HMT-Hydro) experiment concentrated on the 0–6-h time frame for the prediction and warning of flash floods primarily through the experimental Flooded Locations and Simulated Hydrographs product suite. This study describes the various model guidance, applications, and evaluations from both testbed experiments during the 23 June 2016 flash flood event. Various model outputs provided a significant precipitation signal that increased the confidence of FFaIR experiment participants to issue a high risk for flash flooding for the region between 1800 UTC 23 June and 0000 UTC 24 June. Experimental flash flood warnings issued during the HMT-Hydro experiment for this event improved the probability of detection and resulted in a 63.8% increase in lead time to 84.2 min. Isolated flash floods in Kentucky demonstrated the potential to reduce the warned area. Participants characterized how different model guidance and analysis products influenced the decision-making process and how the experimental products can help shape future national and local flash flood operations.

Pattern-Based Decoding for Wi-Fi Backscatter Communication of Passive Sensors

Ambient backscatter communication enables passive sensors to convey sensing data on ambient RF signals in the air at ultralow power consumption. To extract data bits from such signals, threshold-based decoding has generally been considered, but suffers against Wi-Fi signals due to severe fluctuation of OFDM signals. In this paper, we propose a pattern-matching-based decoding algorithm for Wi-Fi backscatter communications. The key idea is the identification of unique patterns of signal samples that arise from the inevitable smoothing of Wi-Fi signals to filter out noisy fluctuation. We provide the mathematical basis of obtaining the pattern of smoothed signal samples as the slope of a line expressed in a closed-form equation. Then, the new decoding algorithm was designed to identify the pattern of received signal samples as a slope rather than classifying their amplitude levels. Thus, it is more robust against signal fluctuation and does not need tricky threshold configuration. Moreover, for even higher reliability, the pattern was identified for a pair of adjacent bits, and the algorithm decodes a bit pair at a time rather than a single bit. We demonstrate via testbed experiments that the proposed algorithm significantly outperforms conventional threshold-based decoding variants in terms of bit error rate for various distances and data rates.