Building on top of the established concept of Mobile Clouds defined as computing in groups of wireless devices at network edge, this paper considers a Smart City in which client devices are expected to aggregate and interact with digital environment of the city in groups (for resource pooling, better performance, higher efficiency, offload, etc.). This paper proposes a 3-party handshake where cloud services are used as intermediary between two client devices. Devices themselves communicate using hash keys, which ensures that both parties remain anonymous to each other. Performance analysis is performed using real mobility traces. Results show that performance drastically improves when the pure peer-to-peer (p2p) approach is enhanced with limited participation by the Smart City infrastructure.
The topic of dense wireless spaces is well known in academic community, with recent research focusing on ways to avoid and/or control wireless congestion. When a large number of independent IoT devices (infrastructure, cars, smart cities, etc.) attempt to define relative positions via radio channels, the problem of low precision is added to that of congestion. On the other hand, ultrasound can provide much higher precision, yet positioning based solely on ultrasound would require too much overhead for multi-party orchestration. This paper proposes a hybrid method named Wifistle, which uses wireless surroundings at each node as context which is broadcast freely over ultrasound. Multiple units of such context data allows for each node to independently infer positions for other nodes on a virtual ring around itself thus, essentially, building a relative positioning map. The method was implemented and tested on Android smartphones with reliable outcome independent of models of devices used for testing.
This paper proposes the novel idea of a packet traffic genome where sequence of bursts and gaps between bursts are encoded as a sequence of genes that represents the respective flow in analysis. This paper shows that, while individual flow genomes are too diverse for a valid analysis, the gene soup representing the total current traffic aggregate exhibits repeated patterns. The repetition can be used to reduce optimizational complexity of management decisions -- this is when a pre-calculated solution can be used for a repeated pattern. Moreover, this paper shows that the genome concept can be helpful for switching at subflow level (e.g. circuits for individual bursts) and can serve as a visualization tool both at the level of individual flows and total aggregate traffic.
Universities traditionally have centralized portals for student information and applications that access it. Access is normally done with web applications using a browser. Smartphone apps, when present, also heavily rely on web APIs. The use of cloud infrastructure, if any, has limited scope and minimal overlap in functionality with the main web portal. This paper considers a scenario, in which data are split into atomic units (per-student, per-class, etc.) and hosted on containers together with the related web applications. This paper argues that, given that most access is done over the web, the migration to containers requires little effort and even leads to simplified application logic. Among numerous advantages of the proposal, this paper focuses on the aspects of dynamic load balancing which is naturally achieved by packing and migrating containers based on current educational activities within the campus.
The main objective of this paper is to show that large-scale sensor data collection is feasible while relying on vehicles to collect and deliver data to a relatively small number of static nodes on the roadside infrastructure. The proposed technology is based on simple 802.11 beacons (as opposed to 802.15 WAVE and 802.11p and others) which are accessible on common smartphones and are comparatively easier to program on commodity 802.11 hardware. While the proposal is not limited to a particular type of collected data, this paper showcases the proposal on smart power grids, where power plants serve as sources of information, vehicles collect the information while passing on the road near the plants, and roadside nodes finally collect the aggregated information from passing vehicles.
The core idea behind Hadoop is to distribute both the data and user software on individual shards within the cluster. The Bigdata Replay method is drastically different in that it packs user software into batches on a single multicore machine and uses circuit emulation to maximize throughout when bringing data shards for replay. The effect from hotspots, defined as drastically higher access frequency to a small portion of (popular) data, is different in the two platforms. This paper models the difference numerically but in a relative form, which makes it possible to compare the two platforms.
Many people in the software development community today do not realize that techniques like SCRUM were originally proposed as tools for general product/project development and problem solving in a traditional (non-specific) business environment. This misunderstanding is mostly due to the fact that SCRUM, Agile, Waterfall and other techniques are used mostly for software development today. Another indicator of this bias is that generic PBL-related certification, namely EXIN and PMP, remain under-proliferated in Japan. This paper shares experience from trying to bridge this gap while teaching PBL at a business school from the prospective of software development. Without ever actually touching software, the well-known SCRUM techniques are generalized to development of any product or service. The paper focuses on several specific techniques and concepts like Backlog, Burndown chart, strategies related to per-task budget allocation, and others.
Cloud services are rapidly expanding to devices at network edge where they are referred to as fog clouds. User smartphones, portable WiFi devices, vehicles, etc., can be easily incorporated into fog clouds. However, proliferation has been slow mainly because the 3G/LTE connections used by such devices to connect to the cloud are congested and cannot provide satisfactory throughput. This problem can be resolved when devices are grouped, their 3G/LTE resources are pooled and their virtual use is optimized within the group. This paper is the first known attempt to propose a minimal secure protocol for forming groups of WiFi IoT devices. The minimal requirement is that the pairwise handshake is a 3-party transaction, with the cloud being both the facilitator of the transaction as well as the only trusted party. The handshake starts at the level of WiFi Access Points (hotspots) but, once paired, the devices are assumed to use P2P WiFi in parallel with the active 3G/LTE connection, i.e. as part of a GroupConnect session. The proposal is implemented and successfully tested on Android devices.
With software defined networking rapidly becoming industry default, the cut-through mode of switching is becoming increasingly harder to achieve in practice. On the other hand, cut-through networking remains firmly in focus in several fields where top network performance is required, where supercomputing and datacenter interconnect are two popular examples. In earlier work, cut-through mode on end-to-end paths was achieved by emulating circuit switching over traditional packet networks. This paper focuses on communication nodes themselves and attempts to solve the practical problem that datacenters encounter when implementing end-to-end circuits. This paper models the problem as an N-by-M switchboard, where N is the number of physical ports and M is the number of ports that compete for the bandwidth on each port in N. The optimization problem is of the bin-packing variety with unique aspects related to the cut-through mode. This paper solves the problem using Genetic Algorithm and proposes an effective heuristic to speed up the optimization.
Recent literature on traffic control in datacenter intra- and interconnect has moved from optical burst switching and FCoE to the cut-through mode enforced e2e on commodity ethernet lines. E2e cut-through sessions -- aka, effectively, circuits -- are implemented as optimized schedules in which bulk transfer sessions are mixed with a large number of small flows. This paper presents a new optimizational viewpoint, formulated as the Switchboard Traffic Engineering Problem (STEP). First, the mixed schedule of input flows is mapped to a finite number of (physical) output lines some of which are implemented as (physical) e2e cut-through paths. The second part of the problem is to physically shuffle input ports (hence the switchboard) on switches with output lines in such a way that overall contention is minimized. This paper deals with the formulation of the optimization problem and leaves the otherwise obvious (robotics) implementation of the shuffling part of the technology to future publications.
Growing number of wireless devices nowadays often results in congestion of wireless channels. In research, this topic is referred to as networking in dense wireless spaces. The literature on the topic shows that the biggest problem is the high number of concurrent sessions to a wireless access point. The obvious solution is to reduce the number of concurrent sessions. This paper proposes a simple method called Bulk-n-Pick which minimizes the number of prolonged concurrent sessions by separating bulk from sync traffic. Aiming at educational applications, under the proposed design web applications would distribute the main bulk of content once at the beginning of a class and then rely on small messages for realtime sync traffic during the class. For realistic performance analysis, this paper first performs real-life experiments with various counts of wireless devices, bulk sizes and levels of sync intensity. Based on the experiment, this paper shows that the proposed Bulk-n-Pick method outperforms the traditional design even when only two concurrent bulk sessions are allowed. The experiment shows that up to 10 concurrent bulk sessions are feasible in practice.
Research on Content Delivery Networks (CDNs) offers a number of effective methods for caching of content replicas or forwarding requests. Clouds, however, present a new kind of distribution environment as each location can support multiple caching options varying in the level of persistence of stored content. This paper proposes a distributed caching technology specific to fog clouds which, by definition, consist of a high number of loosely controlled heterogeneous locations. To deal with the heterogeneity problem and, at the same time, avoid centralized control, this paper proposes a function that allows CDN applications to discover local caching facilities dynamically, at runtime. For simplicity, this paper analyzes the case when each location offers two caching options: VM-based for each app and location-global shared by all local apps. Analysis in this paper shows that the new method can reduce inter-cloud traffic by between 16 and 18 times while retaining less than 30% of total content in a local cache.
The packets versus circuits argument is several decades old. Back in the day, what we today refer to as all-IP networking was justified for its efficiency over the circuit-based telephone networks. The analysis was performed on ON-OFF sources for voice (phone calls) and packets came out to be almost twice as efficient. Modern traffic follows the hotspot distribution, where the hotspots -- also known as elephant flows, bulk transfer, etc. -- have higher value than other flows and therefore require higher efficiency. The highest possible efficiency for bulk transfer can be achieved using circuit switching, implemented on IP networks via the cut-through switching mode. This paper proposes a new switching logic which assumes that NOC has both packet and circuit lines and strives to optimize the newly defined efficiency accordingly.
Roadside infrastructure is traditionally assumed to be based on IEEE 802.15 which defines the various opportunistic and cognitive elements necessary for vehicle-specific mobility. However recent advances in wireless technology -- specifically, the fact the Beacons have become highly accessible both on commodity devices and in readily available software -- opens the path towards minimizing the size of roadside infrastructure while also simplifying individual nodes. This paper discusses a case when relatively fewer roadside network nodes are put to the task of having to collect image data from a much larger number of roadside cameras. Collection is happening via vehicles which are used as sync packets between camera and network nodes. Communication is happening solely via beacons using beacon stuffing when necessary. Special attention is paid to short-term communication and contention when wireless density is high.
Metromaps are recognized well in visualization research but are rarely found in applied technologies. Earlier works showcased metromaps as a valid tool for human-robot hybrid learning when mining Big Data. This paper goes one step further and shows that metromaps are good for controlling complexity in search/state space. To accomplish this, any generic context is represented as train lines and stations, where stations can be shared by one or more train lines. Complexity is controlled by focusing on a given station and defining resolution in terms of hop-length on e2e paths to other stations in the overall metromap.
NGN statement on e2e network performance froze in 2008 with the Y.2173 (former Y.mnm) standard which specified a method for coordination across domains, involving active probing, parsing and sharing results, etc. Today, it is widely accepted that future network functions is to be virtualized and controlled by software. However, virtual switches -- the best models of which to the day are based on MiniOS/ClickOS/ClickRouter designs -- mimic the blackboxing policy found in clouds. This paper shows that if this policy is revoked in favor of declarative coordination, complexity for optimization of overlaying service networks can be substantially reduced. The proposed case is represented by domains declaring their resources/capabilities (e.g. e2e throughput and delay) openly, thus allowing service networks to coordinate among each other. Traditional approach is represented by the VNE optimization problem.
3-Way Scripts are simple code units which can automatically respond to (1) class instantiation and function calls, (2) calls on command line, (3) web API calls. The key merit is that all the three methods are handled by the same code without any additional overhead. In additional to the core idea of 3-way scripting, this paper also adds a security feature for the web API function in form of generation and handling of keys across multiple instances of code running at multiple locations. The main objective for the security features is the same as core idea -- to minimize the coding overhead and maximize simplicity and flexibility. The proposed design is showcased as a practical solution for cloud applications which consist of a large number of independent and loosely coupled instances of the same code. In current practice, this showcase is close in nature to popular scale-out platforms like Docker and Heroku.
Distributed and/or parallel code is normally based on elaborate platforms. The main problems with such platforms are (1) constraints placed on operation of the code and (2) overhead imposed by the platform that arbitrates among multiple instances within the running code. This paper argues in favor of platform-less distribution of code. The base unit is referred to as 3-way script, where the three ways are (1) calling a method/function of an instantiated class, (2) executing the code from the command line, and (3) calling a method/function using HTTP requests to a remote web API. The key merit of the proposal is that all the three uses are possible on the same code, which by developer only one -- this code is referred to as a 3-way script. This paper discusses examples of the code written in PHP, while the same design is possible in several other popular programming languages.
Social Robotics normally assumes visual feedback between robotic trainees and human trainers. Given that robots rarely have adequate visual perception/recognition, such systems are noisy and prone to judgment errors. One way to resolve this problem is to simplify the communication channel between humans and robots. This paper uses simple gravity+motion XYZV sensors ubiquitous in modern personal devices -- smartphones in particular -- to power gait-based exo-systems (power leg assist, etc). This paper discusses current work in progress on this topic, specifically (1) gait modeling and recognition of kick-in moments for hardware, (2) use of the XYZV channel in both directions, allowing humans to send feedback in realtime, (3) social robotics constructs and methods that support maximum flexibility in applications of the otherwise traditionally narrow-purpose hardware systems.
Previous research has solidified the core idea referred to as circuit emulation. The chief premise is that bulk transfer -- Big Data shards, for a practical example -- can achieve the best e2e throughput when all the switches along the path are in the cut-through mode, as opposed to the store-and-forward mode when under any form of contention. Most networking technology in the world today, including network virtualization (SDN, NFV, etc.) runs under contention. This paper looks into physical and logical network designs that provide a degree of control over contention which, in turn, creates room for efficient scheduling of e2e circuits. The unit of new mode of networking is a ring, but -- rather than the comeback of Token Rings, this paper will talk about partially overlapping asynchronous rings.
セル生産方式におけるロボットの活用には様々な問題があるが,その一つとして 3 体以上の物体の組み立てが挙げられる.一般に,複数物体を同時に組み立てる際は,対象の部品をそれぞれロボットアームまたは治具でそれぞれ独立に保持することで組み立てを遂行すると考えられる.ただし,この方法ではロボットアームや治具を部品数と同じ数だけ必要とし,部品数が多いほどコスト面や設置スペースの関係で無駄が多くなる.この課題に対して音𣷓らは組み立て対象物に働く接触力等の解析により,治具等で固定されていない対象物が組み立て作業中に運動しにくい状態となる条件を求めた.すなわち,環境中の非把持対象物のロバスト性を考慮して,組み立て作業条件を検討している.本研究ではこの方策に基づいて,複数物体の組み立て作業を単腕マニピュレータで実行することを目的とする.このとき,対象物のロバスト性を考慮することで,仮組状態の複数物体を同時に扱う手法を提案する.作業対象としてパイプジョイントの組み立てを挙げ,簡易な道具を用いることで単腕マニピュレータで複数物体を同時に把持できることを示す.さらに,作業成功率の向上のために RGB-D カメラを用いた物体の位置検出に基づくロボット制御及び動作計画を実装する.
This paper discusses assembly operations using a single manipulator and a parallel gripper to simultaneously
grasp multiple objects and hold the group of temporarily assembled objects. Multiple robots and jigs generally operate
assembly tasks by constraining the target objects mechanically or geometrically to prevent them from moving. It is
necessary to analyze the physical interaction between the objects for such constraints to achieve the tasks with a single
gripper. In this paper, we focus on assembling pipe joints as an example and discuss constraining the motion of the
objects. Our demonstration shows that a simple tool can facilitate holding multiple objects with a single gripper.
【DLゼミ】XFeat: Accelerated Features for Lightweight Image Matchingharmonylab
公開URL:https://arxiv.org/pdf/2404.19174
出典:Guilherme Potje, Felipe Cadar, Andre Araujo, Renato Martins, Erickson R. ascimento: XFeat: Accelerated Features for Lightweight Image Matching, Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)
概要:リソース効率に優れた特徴点マッチングのための軽量なアーキテクチャ「XFeat(Accelerated Features)」を提案します。手法は、局所的な特徴点の検出、抽出、マッチングのための畳み込みニューラルネットワークの基本的な設計を再検討します。特に、リソースが限られたデバイス向けに迅速かつ堅牢なアルゴリズムが必要とされるため、解像度を可能な限り高く保ちながら、ネットワークのチャネル数を制限します。さらに、スパース下でのマッチングを選択できる設計となっており、ナビゲーションやARなどのアプリケーションに適しています。XFeatは、高速かつ同等以上の精度を実現し、一般的なラップトップのCPU上でリアルタイムで動作します。
3. .
教科書のクラウド
SaaS: Software As A Service
PaaS: Platform As A Service
IaaS: Infrastructure As A Service
...
M.Zhanikeev -- maratishe@gmail.com ハイブリッドクラウドとheroku/EC2決算法-- bit.do/141011 3/44
.
3/44