Search for: All records

In this work, we present our attempt to tackle the last-hundred-feet problem for autonomous drone delivery.We take a computer-vision based approach to progressively landing towards a convenient and safe drop-off point at all times (here, at the front/garage door). Specifically, we develop structural semantic segmentation (SSS), a new technique that leverages a single-family house structure to streamline and enhance semantic segmentation in the drop-to-door problem context.We implement SSS into an Android app; Our preliminary evaluation in a residential zone shows SSS is promising to make autonomous drop-to-door in real-time, with no need to wait for slow visual processing. Video demo is available at Youtube [5]. App is released at Github [6]. 

In this work, we have conducted a measurement study with three US operators to reveal three types of problematic failure handling on secondary radio access which have not been reported before. Compared to primary radio access failures, secondary radio access failures do not hurt radio access availability but significantly impact data performance, particularly when 5G is used as secondary radio access to boost throughput. Improper failure handling results in significant throughput loss, which is unnecessary in most instances. Datasets are available at https:// github.com/mssn/ scgfailure. 

Carrier aggregation (CA) is an important component technology in 5G and beyond. It aggregates multiple spectrum fragments to serve a mobile device. However, the current CA suffers under both high mobility and increased spectrum space. The limitations are rooted in its sequential, cell-by-cell operations. In this work, we propose CA++, which departs from the current paradigm and explores a group-based design scheme. We thus propose new algorithms that enable concurrent channel inference by measuring one or few cells but inferring all, while minimizing measurement cost via set cover approximations. Our evaluations have confirmed the effectiveness of CA++. Our solution can also be adapted to fit in the current 5G OFDM PHY and the 3GPP framework. 

In this paper, we study an important, yet unexplored problem of configuration dependencies in 5G/4.5G radio resource control (RRC). Different from the previous studies in 3G/4G networks, 5G/4.5G allows more than one cells to serve a mobile device, resulting in more configuration dynamics and complexity that vary with all the serving cells. We analyze inter-dependency among configurations, categorize dependent misconfigurations, uncover their root causes, and quantify negative performance impacts. Specifically, we formulate configuration updates into a delta state machine (DSM) and unveil two types of dependent misconfigurations among states (inter-state) and within a state (intra-state); They stem from structural dependency and cross-parameter dependency. We further show that such misconfigurations incur service disruption and performance degradation. Our findings have been largely validated with three US operators and one Chinese operator; Our study has uncovered 644 instances of problematic dependencies.

 

Deep neural network (DNN) is the de facto standard for running a variety of computer vision applications over mobile and embedded systems. Prior to deployment, a DNN is specialized by training to fit the target use scenario (depending on computing power and visual data input). To handle its costly training and meet diverse deployment needs, a “Train Once, Deploy Everywhere” paradigm has been recently proposed by training one super-network and selecting one out of many sub-networks (part of the super-network) for the target scenario; This empowers efficient DNN deployment at low training cost (training once). However, the existing studies tackle some deployment factors like computing power and source data but largely overlook the impact of their runtime dynamics (say, time-varying visual contents and GPU/CPU workloads). In this work, we propose OPA to cover all these deployment factors, particularly those along with runtime dynamics in visual data contents and computing resources. To quickly and accurately learn which sub-network runs “best” in the dynamic deployment scenario, we devise a “One-Predict-All” approach with no need to run all the candidate sub-networks. Instead, we first develop a shallow sub-network to test the water and then use its test results to predict the performance of all other deeper sub-networks. We have implemented and evaluated OPA. Compared to the state-of-the-art, OPA has achieved up to 26% higher Top-1 accuracy for a given latency requirement. 

Edge-assisted video analytics is gaining momentum. In this work, we tackle an important problem to compress video content live streamed from the device to the edge without scarifying accuracy and timeliness of its video analytics. We find that on-device processing can be tuned over a larger configuration space for more video compression, which was largely overlooked. Inspired by our pilot study, we design VPPlus to fulfill the potentials to compress the video as much as we can, while preserving analytical accuracy. VPPlus incorporates two core modules – offline profiling and online adaptation – to generate proper feedback automatically and quickly to tune on-device processing. We validate the effectiveness and efficiency of VPPlususing five object detection tasks over two popular datasets; VPPlus outperforms the state-of-art approaches in almost all the cases.