Artificial intelligence (AI) is increasingly becoming crucial to obtain actionable insights from the substantial amount of data generated by the Internet of Things (IoT) networks to optimize operations and increase their efficiency. This article discusses the importance, benefits and challenges, applications, and recent developments of AI in IoT.
Image credit: Guitar photographer/Shutterstock
Importance of AI in IoT
IoT is primarily a network of different devices connected over the Internet that exchange and collect data. The convergence of IoT and AI is transforming the traditional way of functioning in businesses, economies, and industries as AI-enabled IoT creates intelligent devices/machines simulating smart behavior and providing support in decision-making with no or little human interference.
In recent years, IoT has witnessed rapid adoption in several applications due to its ability to collect a substantial amount of data from different sources in real-time. However, the huge amount of data generated by numerous IoT devices increases the complexities of collecting, analyzing, and processing the data.
The use of AI in IoT can address this issue by efficiently managing and analyzing the data and quickly extracting meaningful insights from it. For instance, machine learning (ML) can identify anomalies and patterns in the smart device-generated information about vibration, sound, air quality, pressure, humidity, and temperature. AI-enabled IoT solutions are also significantly more accurate and faster compared to threshold-based monitoring systems.
Moreover, AI in IoT can balance centralized and localized intelligence requirements, ensure effective security against cyberattacks, and balance personalization with data privacy and confidentiality. ML enables the IoT devices to learn from their experience and data to make decisions.
AI-enabled IoT Benefits and Challenges
Improved Operational Efficiency: AI in IoT can analyze the constant data streams obtained from IoT devices and identify patterns undetectable on simple gauges. Moreover, ML can be used to predict operational conditions and identify parameters that must be modified to ensure better outcomes.
Thus, the AI-enabled intelligent IoT can offer insights into processes that are time-consuming and/or redundant and tasks that can be fine-tuned to increase operational efficiency. Companies like Google already use AI in IoT to reduce data center cooling costs.
Better Management of Risks: Companies can predict and understand different risks and ensure prompt response through automation by pairing AI with IoT to manage better financial losses, cyber threats, and employee safety. For instance, Fujitsu utilizes AI to analyze data from connected wearable devices to ensure worker safety.
Development of Improved Services and Products: The combination of AI and IoT can improve existing/create new services and products by enabling businesses to rapidly analyze and process IoT device-generated data. Companies such as Rolls Royce aim to leverage AI while implementing IoT-enabled aircraft engine maintenance amenities for identifying patterns and operational insights in the data.
Elimination of Expensive Unplanned Downtime: In several sectors, such as industrial manufacturing and offshore oil and gas production, equipment breakdown can lead to expensive unplanned downtime.
Predictive maintenance using AI-enabled IoT can enable businesses to make equipment failure predictions in advance and schedule maintenance procedures in an orderly manner to prevent the adverse effects of downtime. AI-enabled IoT can also reduce the time spent in maintenance planning, increase uptime and equipment availability, and decrease maintenance costs.
Flexible automation: A highly flexible automation can be realized by integrating ML and IoT. For instance, IoT-connected robots with AI functionality that communicate with each other can properly interpret their workflow data, including unusual situations encountered, to identify ways to adapt to changing situations, resulting in better handling of disruptions and more practicality in automation.
Faster Analytics: Smart/AI-enabled IoT devices can analyze their information by themselves without sending the data to a data center located at another location, leading to lower latencies and faster speeds that are beneficial in several applications.
For instance, lower latencies can enable self-driving vehicles to identify obstacles within a fraction of a second, resulting in safe navigation. Similarly, supply chain leaders can obtain information on impending supply chain disruptions in advance due to faster analytics, which accelerates the decision-making process to reduce the impacts of such disruptions.
Improved Cybersecurity: AI algorithms can monitor unauthorized access or suspicious activity actively on IoT devices. Although several companies are already using AI to constantly monitor IoT networks, the use of AI in IoT devices brings this monitoring process closer to endpoints to mitigate potential damages.
Although AI-enabled IoT has several benefits, the integration of these technologies also leads to multiple challenges. For instance, ensuring data security and privacy is extremely challenging in the AI-enabled IoT ecosystem as both technologies collect an extremely large amount of sensitive user data, and the data is typically stored in the cloud.
Compatibility is another significant challenge as several devices possessing different technologies are connected in an IoT network, which can increase the complexity of AI integration. Artificial stupidity, which implies the inability of an AI program to properly execute basic tasks, must be avoided by developing robust and effective AI algorithms to ensure appropriate understanding and interpretation of IoT device-generated data.
AI-enabled IoT Applications
AI-enabled IoT can play a crucial role in home automation, oilfield production, smart hotels, manufacturing, self-driving cars, retail analytics, smart thermostat solutions, body trackers, security devices, smart cities and traffic management, and smart finance/entertainment/health.
For instance, an AI-IoT-based control and monitoring system for home automation can ensure intelligent energy preservation by automating and controlling most electrical appliances, such as fans and lights, through a manageable smartphone-based Android interface. Similarly, AI-IoT can optimize oilfield production by analyzing the data obtained from sensors that measure well pressure, oil extraction rates, and temperature.
Smart hotels using AI-enabled IoT can provide room temperature control flexibility, smart booking system, useful information selection based on customers, real-time assistance to customers on online platforms to address their issues, and re-synchronize customer history by returning guests.
In retail operations, AI-IoT can increase operational efficiency by real-time tracking of inventory levels, improve the store journey experience of customers by increasing engagement through devices such as smart mirrors, increase sales, and effectively forecast consumer demand by analyzing customer data to better understand consumer behavior.
Smart thermostat solutions can manage and measure temperature from any location depending on the user's temperature preferences and work schedules. AI-powered IoT devices in self-driving cars can predict pedestrian behavior in different situations by determining weather, optimal speed, and road conditions.
Moreover, the application of AI-IoT for smart city management can significantly reduce maintenance and infrastructure costs by analyzing the data collected by IoT devices installed around the city to identify the most appropriate solutions to existing problems using minimal resources.
Several real-world IoT services, such as voice assistants, robots, and smart devices, have integrated AI techniques with them. For instance, voice assistants such as Alexa, Siri, and Google Assistant have integrated AI techniques such as conversational AI, dialogue management, contextual reasoning, speech-to-text conversion, automatic far-field voice recognition, natural language processing and understanding, and wake word detection.
Recent Developments
The integration of AI into the existing IoT application programs requires significant coding effort. In an article published in IET Networks, researchers proposed a solution, designated as AItalk, to address this issue. In this novel approach of AItalk, the ML mechanism was considered as a cyber IoT device, which differed from the conventional AI-based IoT applications that fully integrate the AI mechanism into the network applications.
The approach allowed the disintegration of a complex AI application into distributed simplified modules connected using IoT technology. Additionally, the approach also enabled real-time simplified data processing for an AI application. Thus, the solution facilitated the seamless integration of ML capability to IoT applications without requiring any programming effort. The study results demonstrated that the IoT communication overhead in AItalk and the AI prediction time were less than 30 ms and two ms, respectively.
In IoT environments, the existing deep learning (DL) systems cannot reasonably assign computing tasks, leading to the wastage of resources. In a paper published in the IEEE Wireless Communications Letters, the authors proposed the accelerating AI in IoT (AAIoT) method to assign the inference computation of every network layer to each device in a multilayer IoT system.
A dynamic programming algorithm was designed to minimize the response time considering the transmission and computation cost. Simulation results displayed that this approach can significantly improve the system response time.
References and Further Reading
Lin, W., Lin, B., Liu, Y. (2019). AItalk: A tutorial to implement AI as IoT devices. IET Networks, 8(3), 195-202. https://doi.org/10.1049/iet-net.2018.5182
Zhou, J., Wang, Y., Ota, K., Dong, M. (2019). AAIoT: Accelerating Artificial Intelligence in IoT Systems. IEEE Wireless Communications Letters, 8, 3, 825-828. https://doi.org/10.1109/LWC.2019.2894703
Ghosh, A., Chakraborty, D., Law, A. (2018). Artificial intelligence in Internet of things. CAAI Transactions on Intelligence Technology, 3(4), 208-218. https://doi.org/10.1049/trit.2018.1008
Artificial Intelligence in IoT [Online] Available at https://data-science-ua.com/industries/ai-in-iot/ (Accessed on 01 September 2023)
AI and IoT Blended - What It Is and Why It Matters? [Online] Available at https://www.clariontech.com/blog/ai-and-iot-blended-what-it-is-and-why-it-matters (Accessed on 01 September 2023)
Tzafestas, S. G. (2018). Synergy of IoT and AI in Modern Society: The Robotics and Automation Case. Robotics & Automation Engineering Journal. https://www.researchgate.net/publication/327837834_Synergy_of_IoT_and_AI_in_Modern_Society_The_Robotics_and_Automation_Case
Esraa, M. (2020). The Relation Of Artificial Intelligence With Internet Of Things: A survey. Journal of Cybersecurity and Information Management (JCIM), 1, 1, 30-34. https://www.researchgate.net/publication/340006839_The_Relation_Of_Artificial_Intelligence_With_Internet_Of_Things_A_survey
Newton, E. (2023). 5 benefits of Artificial Intelligence IoT. [Online] Available at https://techinformed.com/benefits-of-artitifical-intelligence-iot/ (Accessed on 01 September 2023)