Home automation systems and networks aim to boost the quality of life and support the automation of industrial operations. Their wide applicability has attracted the interest of researchers and companies. In the last decades, home automation systems constitute an emerging domain, and the corresponding developed technologies are available to both organizations and infrastructures as well as common users. The importance of home automation systems is apparent as their technologies are utilized in various important domains such as smart homes, Internet of Things (IOT), vehicles and healthcare. Companies that develop products of this nature aim to patent their most valuable technologies in order to protect their investments against rival competitors. The granted patents are recorded and stored by different patent offices that provide comprehensive information and constitute the knowledge base in much research. In this study, we analyze patent records from the United States Patent and Trademark Office (USPTO) that are related to home automation networks in order to reveal the most relevant technologies and domains of application as well as the respective prominent stakeholders. Our findings provide insights on the current state of home automation technologies and support researchers and organization in tasks that are related to competitor analysis, strategic planning and technology forecasting.
Ubiquitous computing is a concept that encompasses multiple different areas of the computer science domain, including distributed computing, sensors and IoT technologies as well as artificial intelligence. A primary aspect of ubiquitous computing is its objective on affecting the daily aspects of human life, providing time-saving and innovative alternatives to otherwise tiring tasks. An example of ubiquitous computing that highlights its presence in the modern society are home automation systems, expressed in devices that range from digital assistants to smart lightbulbs and devices that can be found in an average household.
The importance and benefits of home
automation systems are more than evident in
everyday activities, as they provide capabilities
that facilitate and improve home security,
management, flexibility and remote control.
The main components that encompass a
complete home automation system are the user
interface, the participating electronical devices
and a mode of transmission [
Given that inventions with innovative
concepts of home automation technologies
provide organizations and companies with
ample opportunities of exploiting them in the
market, the early patenting of inventions has
become a staple for investors of not only home
automation, but also in any other technological
domain. Thus, patent data can be a potent
indicator of technological growth, development
and business activity [
In this study, we explore the main applications and technologies that are related to home automation systems through a patent analysis methodology that employs dimensionality reduction and cluster analysis techniques. To do so, we collect and analyze information of technological patents, derived from the USPTO, that are related to home automation systems. Complementary, we also reveal the main stakeholders that invest in home automation technologies and contribute to the landscape profiling of the investigated technological area and subareas.
The results of this study provide insights for both practitioners and industries in identifying the main domains, applications, methodologies and competitors of home automation technologies. In the context of disseminating knowledge, this study serves as a steppingstone in merging the disciplines of ubiquitous computing and patent analysis, offering valuable information that can foster and promote additional research on the topic. In addition, our methodologies constitute a robust example of applied Machine Learning on ubiquitous computing data, indicating breakthrough findings related to patents of the field via a streamlined data analytics pipeline.
Patent analysis is a process that is related to
analyzing information of patent data and
2
https://www.uspto.gov/web/patents/classification/cpc/html/cpc.h
tml
extracting knowledge for various purposes
including trending analysis, technology
forecasting, strategic positioning and
competitor analysis [
Patent analysis is a well-known technique
that has application in different technological
domains that are related to augmented reality
[
The first goal of this study is to discover the general technological areas where home automation networks are applied and the methodologies that are leveraged to construct this type of networks. The second goal is to map the main assignees of technological patents, that are related to home automation networks, and identify the main stakeholders of each technological area, as measured by the overall patents that they possess. These two goals are summarized into the following Research Questions (RQs):
RQ1: What are the main technological areas that are related to home automation networking technologies?
RQ2: Who are the main stakeholders of the different technologies that are related to home automation networking?
The main framework of this study is presented in Figure 1. 3 https://www.wipo.int/classifications/ipc/en/
Data Collection
K-means
RQ2 PCM
PCA Main Stakeholders
RQ1
Granted Patents 800 600 400 200 0
In order to discover the general areas of the
investigated domain, we employ a
methodology similar to [
In this approach the first step is to construct the Patent CPC Matrix (PCM) as follows (1). 1, when ℎ − ℎ (1) =
ℎ − ℎ
{ 0, ℎ where i=1,2,…,N and j=1,2,…,M. N is equal to the number of the collected patents, 5740 in our case, and M is equal to the number of the distinct CPC subgroup IDs assigned to the collected patents (5523 overall).
In the next step, since the number of CPC
subgroups is quite extensive, we apply the
Principal Components Analysis – PCA [
K-means
clustering algorithm [
NPMI [
The silhouette coefficient of a single data point is presented in (ii) (2) s(i) =
( )− ( ) max ( ( ), ( )) where ( ) is the mean distance between the i-th data point and the points that belong to the same cluster while ( ) denotes the minimum mean distance between the i-th data point and the points of the remaining clusters.
While the silhouette coefficient is a standard approach to evaluate clustering models, the NPMI will help us to interpret the extracted clusters in terms of general technological areas using the five most probable CPC subgroup IDs of each cluster. The NPMI between two features is presented in (3). (2) (3) where ( ) denotes the probability of a patent being assigned to the i-th CPC subgroup identifier and ( , ) denotes the probability of a patent being assigned to both i-th and j-th
Finally, we make use of the information related to the assignees of each patent, included in the patent data, to reveal the main stakeholders of each technological area. In this phase, we rank the linkage strength of an assignee and a cluster according to the number of patents and belonging to this cluster and under the ownership of this assignee (Main Stakeholders).
4.1.
RQ1:
What are the main technological areas that are related to home automation networking technologies?
By leveraging the patent properties extracted from the PCA, we conduct several experiments to choose the number of clusters that optimize the NPMI and the silhouette coefficient. In this study, we set the range clusters from 2 to 20. The evaluation of the extracted models is presented in Figure 3.
10No_Clusters 5 15 20 of the collected patents belong to the tenth cluster that represents the general area of home automation networks as multiple patents are assigned only to the CPC subgroups that are related to the prefixes of Table 1. By inspecting the most probable CPC subgroups, we can conclude that the results do not indicate significant overlaps between clusters apart from some cases where the subgroups of the last cluster occur in other clusters as well. As the last cluster is related to the general investigated technological area, the identified overlaps should be characterized as insignificant and expected.
This observation shows that the applied approach discovered distinct and interpretable technological areas. These areas are related to both interesting and established techniques applied on home automation networks e.g., Error detection and Information Security, Information retrieval and transferring and important domains that leverage technologies of home automation networks e.g., healthcare, vehicles, Storage and control devices, air screening, games.
In this subsection we present the main assignees contained in the retrieved patents. The involvement of the investigated assignees reveals their overall interest in the technological area of home automation networks. This involvement can be translated on investments in patent technologies, their desire for applications of these technologies and the strategic advantages or disadvantages against competitors in the technological area of home automation networks. In brief, a total of 922 assignees are detected in the retrieved patent data with more than 50% percent owning a single patent, while the assignees’ mean number of granted patents is less than 7 and the largest number of granted patents is 422, owned by Samsung. This statistic indicates that the majority of assignees are not investing in many home automation network patents, being overrun by competitive technological giants, while there is also a small proportion of assignees that invest frequently in technologies that are related to home automation networks, while having the resources and market shares to do so.
By combining the information of the patent data with the technology clusters that were extracted in the previous subsection, we are able to distinguish the main of assignees of each cluster. The top three most involved assignees of each technology cluster are presented in Table .
Table 2 Main assignees of the technology clusters
Top assignees per cluster STATE FARM MUTUAL AUTOMOBILE INSURANCE COMPANY ; 16Lab Inc ;
2Wire, Inc. (Cluster 1) WINT WI Ltd ; 16Lab Inc ; 2Wire, Inc.
(Cluster 2) May Patents Ltd. ; 16Lab Inc ; 2Wire,
Inc. (Cluster 3) Cisco Technology, Inc. ; 16Lab Inc ;
2Wire, Inc. (Cluster 4)
University of Florida Research Foundation, Inc. ; Delos Living LLC ;
16Lab Inc (Cluster 5) Microsoft Technology Licensing, LLC ; Canon Kabushiki Kaisha ; Hitachi,
Ltd. (Cluster 6) Intel Corporation ; 16Lab Inc ; 2Wire,
Inc. (Cluster 7) Broadcom Corporation ; 16Lab Inc ;
2Wire, Inc. (Cluster 8) Google Inc. ; 16Lab Inc ; 2Wire, Inc.
(Cluster 9) Samsung Electronics Co., Ltd. ; Sony
Corporation ; Google Inc. (Cluster
10)
By inspecting the results presented on the table above, we detect that 16Lab Inc and 2Wire, Inc. are involved in the majority of the extracted technology clusters. 2Wire, Inc.4 was a home networking equipment manufacturer that provided products to telecommunication companies while 16Lab Inc5 provides an integrated platform for IOT and wearable developers. While these two assignees are strongly related to home automation networks, we also observe some established software and hardware companies, that do not only invest in home automation networks but lead pioneering developments in other fields as well, like Google, Samsung, Sony, Microsoft and Cisco.
Among the different areas of interest of the assignees, we can distinguish that farming, automobiles, information security, IOT, hardware, software development, healthcare and telecommunications are the most frequent ones. These observations reveal that home automation networks are useful in many technological areas, systems, devices and applications.
In this study, we presented and uncovered the main technological areas and assignees that are related to home automation networking by analyzing patent data derived from the USPTO. The first step in our approach was to investigate the patent objectives using the CPC assignments that characterize them. Next, by applying a dimensionality reduction technique and a clustering algorithm, we succeeded in identifying and interpreting the main technology clusters of home automation network patents, providing effective answers to RQ1.
Moreover, the information regarding the assignees of each patent and the technology clusters extracted from the K-means algorithm help us in identifying the main stakeholders of the different technologies that are related to home automation networks (RQ2).
Our approach unveils some interesting technological areas that are related to home automation networking which support the utilities of farming, healthcare, automobile, security and information systems. Furthermore, the applied methodology reveals the main investors in each domain that applies home networking technologies and their interest in patent granting.
The challenges that arise from the combination of ubiquitous computing in machine learning stem from the proper understanding of the field and its usage on daily life as well as the identification of proper datasets for analysis. Nevertheless, we believe that our insights prove that such challenges can be overcome when studying ubiquitous computing from an industrial perspective.
By recapping, we believe that the proposed approach provides valuable information that could be useful for companies and individuals in scoping the main competitors and potential collaborators in future projects. We expect that this information can also provide guidelines to practitioners and researchers in identifying important technologies and methodologies in the general domain of home automation 4 https://en.wikipedia.org/wiki/2Wire 5 https://16lab.net/about/ networking, which could be a great asset in future studies and in the forging of new inventions and applications. As far as the development and knowledge sharing on ubiquitous computing is concerned, this study is a great example of interdisciplinary research that promotes complementarity in science, a concept that should be leveraged by other researchers.
Also, we believe that the proposed methodology can provide effective results not only on patent data but also on research literature data. For instance, the CPC subgroup identifiers in the clustering approach could be replaced with the keywords of each study while the patent assignees could be replaced with the authors and publication venues of the studies. In general, this approach can be used to identify the main individuals or contributors of a dataset and further discover the main areas or topics of that are reflected on the dataset records.
Evidently, there were some limitations in our study. Firstly, the data that were used for this study are derived exclusively from the USPTO and not from all the available patent offices. Additional data from the various related industries and patent offices should be evaluated in order to generalize the results of this study. Despite the occurrence of these limitations, USPTO is characterized as a reliable source for patent analysis that reflects on the industrial technologies and interests.
Furthermore, additional experiments should be conducted in order to verify our approach as an effective alternative for general tasks and different data. Some important stages regarding the evaluation and interpretation of the clusters were conducted in a manual fashion. As a result, we believe that several alternatives in different stages of this approach should also be investigated. Despite the existence of the issues described previously, the algorithms and evaluation metrics of the proposed approach are similar to multiple existing studies that provide satisfying methodologies and outcomes.
Nonetheless, we consider that future studies could also explore the general domain of home automation networks through additional information and purposes. The patent titles and abstracts are two very important features that describe the nature of the inventions and could possibly be analyzed for knowledge discovery through topic modelling or co-word analysis. Also, the patent granting dates, filing dates and inventors are features that are included in the patent data and provide useful information about the overall temporal growth of a technological domain and the most prolific individual inventors respectively. Finally, we believe that the citation properties of the patents contain valuable information regarding the technological value and influence of the patents that help in identifying emerging and dominant technologies and determining the strategic positioning of assignees.