Supervised learning− Unsupervised learning− Semi-supervised learning− Reinforcement learning
Agent: Refers to something that perceives the environment through sensors and performs action via actuators based on some predefined rules for which the agent is trained. Just like in humans, there are five sense (i.e., sight, sound, smell, taste, and touch), and based on the information of our senses we perform actions through our limbs, etc. A simple diagrammatic representation of an agent is shown in Figure 2.2 below.The agents act as a backbone for AI techniques that govern how they are working and what sorts of applications they are dealing with. Based on utility there are different types of agents [10]:− Simple reflex agent− Model-based agent− Goal-based agent− Utility-based agentFigure 2.2 Representation of an agent.
Internet of Things: Refers to the interconnection of the things that we use in our daily lives with the internet. The basic idea of the IoT [11] is to connect all the devices through the internet via short-range wireless devices, such as Zigbee, Bluetooth, RFID, and various sorts of sensors and devices [12], by which they can communicate with each other and share the sensor data among the peer devices in order to facilitate the end-user/client via cloud services. It is one of the most demanding technologies in any industry and from 2019 to 2025 it is expected to grow by 33.81% [13]. The basic architecture of the IoT is shown in Figure 2.3 below.
Robots: Programmable machines capable of performing complex tasks in intense and rigorous environments. They are designed as in Figure 2.4 to automate any human tasks with the controllers built-in or outside based on the requirements. Robots are now being used in many places [14] to perform dangerous or repetitive tasks to safeguard humans, and are being used especially in industry in painting jobs, warehouses, assembly lines, etc. So, basically, they are automating the environment in order to reduce human efforts. And now robots have become an essential commodity of many sectors; for example, they find application in healthcare [15], education, research and development, and are used in architecture, as waiters in restaurants, and almost all the sectors of our economy. With the increase in research and innovation, more and more intelligent systems are being prepared that are forming the foundation of modern society and helping in reshaping the future.
Figure 2.3 Architecture of the Internet of Things.
Figure 2.4 Basic robot architecture.
2.4 Requirements of AI and IoT for Robotic Automation
Robotic automation is one of the most challenging and lengthy tasks without the use of any proper tools and techniques and today that requirement is fulfilled by one of the most advanced techniques available to us, so-called Artificial Intelligence and the Internet of Things, with the help of which we are able to design more complex and more powerful robots. For instance, if we look at the very first form of computer and the computers now available to us, i.e., powerful computers in our pockets, we can see a huge difference. In the same way, the technologies are transforming our lives and giving birth to more advanced forms of machinery that we had once only thought possible.
So, while AI and IoT are required because they are bundled with full-fledged tools and techniques, which are enough to make a powerful robot using traditional programming, it will take months writing the same codebase in comparison to the codes that are written using these technologies. Basically, AI and IoT are efficient and effective enough to work in developing any robotic system.
2.5 Role of AI and IoT in Robotics
Artificial intelligence (AI) and Internet of Things (IoT) are the technologies of today and they are becoming more and more advanced day by day. These technologies are in very high demand in the industry as all the innovations taking place are based on them. AI comprises mathematical and statistical models that govern the working of algorithms that are used to develop intelligent systems and the IoT consists of various tools and techniques to effectively manage the sensors and their intercommunications via the use of various protocols and devices. Due to the high capabilities of these technologies, they are highly adopted in industries, healthcare, businesses, and various sectors of the economy [16].
Both of these technologies, when used together, can be much more beneficial as IoT is better at collecting data and AI is a great tool to process huge amounts of data. As IoT uses other technologies like big data or AI for data processing, this implies that AI works on the backend of IoT and plays a major role in working on any system or framework comprising the two [17]. The perfect example can be our voice user interface devices such as Alexa or Google Home. They were trained with some data and that data has been processed via AI whose engine gives output on the basis of data.
Similarly, in robotics, which is a complex system consisting of various types of sensors, various electrical and mechanical devices work together to perform an assigned task. In robotics for the case of recognizing and classifying tasks, it uses computer vision in which the thousands of raw images of objects are fed into machines, and once trained it can classify and recognize objects. In this case, the camera will capture the images from the surrounding (which falls in the IoT domain) and gives it to the ML engine for processing, and once processed the output is shown via actuators or via any output devices.
Some of the major roles of AI and IoT in robotics are:
To program various aspects like learning, understanding, thinking, and inferring based on rules into the robot so as to perform accordingly.
To implement various supervised, unsupervised, semi-supervised, or reinforcement learning algorithms into the robots based on the utility of robots either in industry, business, or for commercial purposes.
To establish various connections between different parts of the robot-like camera connection, wireless modules like Zigbee or Bluetooth, connections between microcontrollers to actuators, etc.
To set up a trained classifier or model so that it can be used by the robot.
To install sensors and actuators so as to sense the environment and perform accordingly.
To create an inference engine for performing inference on the basis of a percept sequence or percept history.
In some cases, to enable speech synthesis so as to talk or control via voice, i.e., voice user interface (VUI) [17].
To establish connections via the cloud so that it can be remotely configured or controlled.
The performance of a robot is governed by its memory or percept sequence and, while training, the model finds some sort of patterns in data that form the basis for learning. While creating a robot there are various aspects that need to be taken care of and the aim is to develop a cognitive architecture in which integration of reasoning, planning, reacting, creating, learning from the past, etc. [18] exists. Inspired by human biology, we try to mimic every biological behavior artificially in robots, like neural networks being built to mimic the behavior of our brain into machines, and various joints being artificially created which were inspired by the human body.
Apart from that, artificial organs are being created to help needy people. Although mimicking the human brain is such a typical task, various companies/researchers are working day and night to build a replacement for the human