Patent information can provide many valuable insights that can be exploited for developing and implementing new technologies/ products/services. Patents also can be exploited to identify new use case development opportunities.

Blockchain-based Privacy-Preserving Federated Learning System:

A federated learning system trains a machine learning model based on data generated from large numbers of users interacting with their devices while the users maintain their data locally in their devices. Each data owner (e.g. federated learning participant) only provides locally trained machine learning model to a trusted third party (aggregator) for a single consolidated and improved global model. The global machine learning model is then sent back to user devices iteratively for updating the local machine learning model. Such collaboration among the federated learning participants lead to more accurate machine learning model than any party could learn in isolation. The federated learning system is a collaborative machine learning method which alleviates privacy issue by performing the machine learning training process in a distributed manner, without the need of centralizing private data.

Federated learning systems, however, provide insufficient data privacy. To protect the privacy of the datasets, federated learning systems need to also consider inferences derived from the machine learning process and information that can be traced back to its source in the resulting trained model. The conventional attempts to ensure adequate data privacy in federated learning systems have resulted in poor predictive performance of the resulting model. For example, federated learning systems using local differential privacy[2] can result in the generation of an abundant amount of noise, which can deteriorate machine model performance. To provide additional measures for privacy protection, blockchain can be integrated to form decentralized federated learning systems.

IBM’s patent application US20200394552 illustrates a blockchain-based federated learning system. In the blockchain-based federated learning system, client nodes of the blockchain network act as federated learning participants. Each training participant client trains a machine learning model using a stochastic gradient descent and its variants. Stochastic gradient descent is a standard optimization technique used in training machine learning models. This technique involves computation of the so called “gradients” of a particular loss function defined on the training dataset and the type of machine learning model involved. In a typical machine learning training process, one starts from an initial machine learning model and updates the model iteratively based on a stochastic gradient descent calculated in each step of iteration. The role of the blockchain-based federated learning system is to ensure that the model is being properly trained by verifying gradients calculated by each federated learning participant at each step of iteration.

Blockchain-based Decentralized IoT & AI Data Marketplace:

Interconnected IoT devices can generate a huge amount of date that can be used for IoT applications.  IoT data marketplace can connect IoT data sellers and buyers so that IoT data can be collected, processed and finally consumed by different parties for IoT applications. Blockchain enables a trusted IoT data marketplace that allows secure and anonymous trading of IoT data. The decentralized nature of the data marketplace based on blockchain means that any participant who qualifies can enter the marketplace as a data seller or a data buyer. Decentralization also means that there is no central authority to regulate the participants of the market. There is no central data repository. The data sellers are the owners of, and remain in full control over, their data. Grandata Inc’s patent application US20200058023 illustrates a blockchain-based decentralized data marketplace that can trade IoT data and other types of data. An example of a data buyer is a company who would like to train its own machine learning models using data purchased via the decentralized data marketplace.

Blockchain-based Decentralized Parallel Edge Machine Learning:

Efficient AI model building requires large volumes of data. While distributed computing has been developed to coordinate large AI computing tasks using multiple computers, applications to large scale machine learning (ML) problems is difficult: For example, in distributed AI computing environments (e.g., Interconnected IoT sensors in a smart factory, Interconnected IoT devices for smart mobility services, connected vehicles), the accessibility of large and sometimes private training datasets across the distributed devices can be prohibitive and changes in topology and scale of the network over time makes coordination and real-time scaling difficult. HPE’s patent application US20190332955 illustrates a blockchain based decentralized machine learning that is performed at blockcahin nodes where local training datasets are generated to build AI models. The blockchain is used to coordinate decentralized machine learning over a series of iterations. For each iteration, a distributed ledger is used to coordinate the nodes. Each blockchain node can participate in a consensus decision to enroll another physical computing node to participate in the first iteration. The consensus decision applies only to the first iteration and cannot register the second physical computing node to participate in subsequent iterations.

Predictive Maintenance Platform for Industrial Machine using Industrial IoT:

Heavy industrial environments, such as environments for large scale manufacturing (e.g., aircraft, ship, automobile manufacturing), energy production environments (e.g., oil and gas plants), energy extraction environments (e.g., mining), construction environments (e.g., construction of large buildings) involve highly complex machines and workflows, in which operators must account for a host of parameters and metrics in order to optimize design, development, deployment, and operation of different technologies in order to improve overall results.

Many of the large industrial machines that require ongoing maintenance, service and repairs are involved in high stakes production processes and other processes, such as energy production, manufacturing, mining, drilling, and transportation, that preferably involve minimal or no interruption. An unanticipated problem, or an extended delay in a service operation that requires a shutdown of a machine that is critical to such a process can cost thousands, or even millions of dollars per day.

IoT enables automatic data collection in industrial environments and AI enables improved methods and systems using collected data to provide improved monitoring, control, intelligent diagnosis of problems and intelligent optimization of operations in various heavy industrial environments. IoT sensors can collect, and AI can process data from industrial machines for predicting faults, anticipating needs for maintenance, and facilitating repairs (predictive maintenance). Blockchain based distributed ledger can record and track maintenance transactions related to the industrial machine. Strong Force IoT’s patent application US20200133257 illustrates a system for detecting operating characteristics of an industrial machine.

5G-based AI+Blockchain+IoT Edge Computing System:

In the coming years, it is expected that there will be a greater need for wirelessly accessible, relatively low-latency, relatively high-powered computing placed near the edge of networks. It is expected that various AI algorithms will need to process relatively high-bandwidth streams of data to output results in real-time after that data is acquired. Examples include processing videos and other sensor data gathered by self-driving cars, autonomous drones, wearable computing devices, and other IoT sensors. In many cases, uploading this data to a traditional public cloud data center to process the data and to generate actionable commands or results is too slow, in part due to the amount of time taken to convey the data over relatively large geographic distances. This is due, in part, to the time consumed transmitting the data and results from the speed of light imposing limits on how fast information can be conveyed over large geographic distances. Additional delays arise from switching and routing equipment along the path and potential congestion. Accordingly, it is expected that there will be a need to distribute relatively high-performance computing facilities, such as data centers, over distributed geographic areas. For example, distributing the data centers every few miles in a metropolitan area, county, state, or country, rather than relying exclusively upon data centers that are geographically concentrated and serve, for example, a continent from a single geographic location.

Edge computing refers to the transition of compute and storage resources closer to endpoint devices in order to reduce application latency, improve service capabilities, and improve compliance with security or data privacy requirements. Edge computing provides a cloud-like distributed service that offers orchestration and management for applications among many types of storage and compute resources. As a result, some implementations of edge computing have been referred to as the edge cloud. Edge computing use cases in 5G mobile network settings have been developed for integration with multi-access edge computing (MEC) approaches, also known as mobile edge computing.

MEC approaches are designed to allow application developers and content providers to access computing capabilities and an IT service environment in dynamic mobile network settings at the edge of the network. MEC technology has some advantages when compared to traditional centralized could computing environments. For example, MEC technology provides a service by service providers to user agent or a user equipment with a lower latency, a lower cost, a higher bandwidth, a closer proximity, and/or an exposure to real-time radio network and context information. Intel’s patent application US20200195495 illtstrates technological solutions for implementing a MEC-based system to realize 5G network slicing with blockchain traceability. The technological solutions integrates MEC with various types of IoT or Fog networking implementations as well as dynamic network slicing and resource utilization management. The technological solutions benefit a variety of use cases, such as 5G network communications for automotive devices, smart factory sensors, and smart healthcare devices.