Evolution of embedded systems

An embedded system traditionally comprises of a microprocessor, microcontroller or a digital signal processor (DSP) as the basic hardware at the core running with a software designed to perform a dedicated function within an external mechanical or electrical system. Around this core structure lie the various functional components comprising a sensor that converts input physical information into an electrical signal, convert electrical (analog) signal to digital (ADC), process digital data and store as reference, convert the processed data into analog signal (DAC), and finally enable the output signal to perform the given task of actuating an external system.

The complexity of the system ranges from a single microprocessor to a suit of processors which may be hosted on readymade embedded boards or system on chip (SOC) implemented with ASICs or FPGA designs, a single user interface to GUI, interfaces for communicating with the external world using USB, Bluetooth, WiFi, Zig Bee, RS 232/485, Ethernet etc, and software written in high level depending on the functionality to perform and then compiled to a code embedded onto the memory to execute, with the architectures varying from single loop to interrupt control to multi-threading and multi-tasking etc.

One of the earliest commercial microprocessors built to use for embedded system was Intel 4004 in 1960s, and the first microcontroller for embedded use was TMS 1000 series by TI in early 1970s. And the first modern real time embedded system built was Apollo Guidance computer also in 1960s. Similarly, the first embedded real time OS was Vx Works.

In these five decades the embedded systems have witnessed tremendous transformation in their capacity and functionality. The drivers for the change in the performance of embedded systems are improved hardware as well as evolution in software. Some of the significant developments can be traced as below:

· Microprocessors going from 4 bit to 8 bit to 16 - 32 bits, with popular Intel 8086 series chips still in use, and lately high-performance CPUs

· Multiple core SOC implemented using ASIC or FPGA designs

· Use of Harvard and Von Neumann architectures for hardware designs with RISC or CISC implementation

· Evolution from single user interface to graphic user interface (GUI)

· Multiple communication interfaces using Bluetooth, Wi-Fi, Zigbee, RS 232/485 etc.

· Evolution of smart sensors with edge computing catering to system mobility

· Tools covering more efficient compilers, assemblers and debuggers

· Systems that are more software defined with sophisticated scale embedded systems using intelligent algorithms

· More recently, using AI algorithms to run on embedded platforms like ARM processor architecture one can achieve a range of complex functionalities like machine vision etc.

· The designs going to higher echelon following the development and use of enterprise systems through virtualization.

Overall, the current day embedded systems enable faster communication, have capability for large data storage for diverse models to run, have highly inter-woven connections among the devices and consume least power and are comparatively of low cost. The core requirements of embedded systems continue to remain the same as below.

· Secure - having protection from being hacked

· Safe - does not affect the application environment and performance is not degraded

· Reliable - performs as expected over its lifetime

· Certifiable - is an important part of the operation

Important features of embedded system

Across a range of embedded systems their features could be summarized as below;

· Customize a combination of hardware, software and firmware designed for specific function

· These are basically low cost, low power consumption small computers that are embedded in external mechanical or electrical system

· These commonly comprise of processor, power supply, memory and communication ports in hardware with software or a firmware running an application

· These may be a standalone system or designed to function within a larger system

· These may be programmable or have fixed functionality

· They may use microprocessor or microcontroller or digital signal processing chips and programmable logic devices as the core hardware

· These may be a computing system with no user interface such as a device for a single task or have a complex graphic user interface such as a mobile phone

. These are often used in a real time operating environment and use RTOS to communicate with the hardware

· User interfaces may be buttons, touch screens, keyboards or, more recently voice commands

· Use computer chips with varying capabilities from 4 bits to 32 bits or multi core SOC implemented with FPGA or ASIC

· Performance wise they vary from small scale (4-8 bit) to medium scale (16-32 bits) to sophisticated scale with large data sets and several algorithms

· Functionalities wise they may fall in the categories of mobile embedded, networked embedded, stand alone, or real time systems

· Often these are used for sensing and real time operating using IOT devices which are internet connected and enable a range of applications

· Simple high volume embedded systems for consumer market are 95% hardware, whereas highly specialized low volume embedded systems would have 95 % software.

Emerging trends in design of embedded systems

Over the years one has witnessed emergence of versatile embedded systems with incredibly increased microprocessor or microcontroller capabilities in processing power, and reduced power consumption and cost.

As one example of the trend, the embedded systems design is changing following the developments in enterprise systems, which has rendered embedded systems more flexible and software oriented, as opposed to the traditional purpose-built embedded systems, unique to each device that run an RTOS Vx Works. Explaining this through an example of an Automobile which may traditionally deploy multiple proprietary embedded systems which work simultaneously, one each for functions like radar, engine control, telematics, connectivity and breaking, each of which has a dedicated microprocessor, memory RAM/ROM, timer, I/O ports and power supply, its own operating system and certification. This approach has been superseded by adopting a software defined open architecture and open standards, leveraging commercially available off the shelf (COTS) devices such as standard computer board, PC platforms etc. to function as the single processing module for execution of the various functions. Such a shift has led to a dramatically reduced cost and faster turnaround. Systems that traditionally worked as isolated are now increasingly connected using the enterprise systems approach. A caveat here is that such systems may be more prone to security breaches as in a connected environment one system can easily provide path to hackers to reach out to other more critical parts of the system. An important  example of such a system is a point of sales (POS) in a connected Retail store. Available techniques in cyber security are utilized to ward off or minimize such eventualities.

Design improvements in embedded system are being leveraged through virtualization- a technique that has been in use with enterprise systems for several years - that allows a single platform of an embedded system to be used to realize its complete functionality, also addressing simultaneously the security, safety, reliability and certification requirements of embedded systems.

The industry of embedded systems continues to grow, driven by accelerated developments with maturing technologies of AI/ML, VR/AR, IOT and associated components. The cognitive embedded systems will be at the core of such trends featured with high performance processing, reduced power, improved security, cloud connectivity and mesh networking. In this category, emerging are GPU powered embedded systems where one would need advanced high-speed analytics platforms to achieve the desired functionality of embedded systems.

AI technology is poised to ingress in a big way in the embedded systems to acquire a certain level of smartness. Here AI is embedded meaning it is encapsulated making an algorithm or model run seamlessly on an embedded system. Citing a simple example, auto turning on or off light in a room depending upon someone is there in the room, or ambient light condition. This application draws from the knowledge of ambient light intensity or physical movements that is acquired as a data from the sensors (in this case by measure of lumens and use of digital cameras) and the use of a model that allows triggering an actuation.

Several such systems are already in development and use in which the embedded systems are equipped with AI that replaces human intervention. Popular examples are in application of image recognition, driverless vehicles, drones, robots for hazardous jobs and jobs requiring precision, retail, and machine vision for Industrie 4.0 eco system. A range of mobile computing devices with integrated software such as mobile/smart phones, smart cameras, smart mobile sensors, tablets, handheld devices with IOT connectivity form a critical part of such embedded systems. In all these examples AI is witnessed as an imperative technology required for the growth and development of embedded systems. In their implementation, often it requires use of massive data that is processed in multiple high-speed computers which may be connected over cloud.

Often enterprise systems use some standard hardware and open-source software through which they quickly build a powerful embedded system which is intelligent and go to market quickly with a cost and performance competitive edge. As an example, machine vision algorithms run on embedded systems- the platforms are called embedded vision. AI based technologies of deep learning (DL) and convolutional neural networks (CNN) are becoming increasingly relevant - more importantly in a highly automated environment- to realize machine vision.

As per the going the future of embedded systems lies in smart use of AI algorithms, ubiquitous computing, edge computing, intelligent mobile (computing) devices, IOT with net connectivity, cyber physical systems (CPS), context sensitive devices, organic computing and the like.

Ubiquitous computing relates to computing that deals with interconnected and communicating devices which are integrated into the objects we interact with. Organic computing implies creating an environment where humans are  surrounded by a variety of autonomous systems, which in turn are equipped with sensors and actuators as part of Cyber Physical Systems (CPS) and communicate among themselves freely and organize in order to perform a given task and offer services as required. CPS as a new family of intelligent engineering systems, built from a close integration of cyber and physical systems, has emerged as a result of fast developments in smart sensors and actuators as part of physical systems, and computing and networking as part of cyber systems.

Applications of embedded systems

From the foregoing discussion it is clear that the embedded systems have found their presence in a wide-ranging application - from consumer to industrial control, smart agriculture, intelligent transportation systems, automobiles, telecom, healthcare and medical systems, airplanes, defense and security, smart grid energy systems, retail, supply chain, smart sensors and devices and more. Some of the popular examples of the usage are;

Telecom - switches, routers, network bridges

Consumer - MP 3, mobiles, digital cameras, digital watches, video games, GPS receivers, personal assistants, vending machines

Households - microwaves, washing machines, dish washers, lighting, heating and air conditioning

Intelligent buildings - security and access control, surveillance, climate control, lighting controls

Transportation - driverless cars, traffic control, situation assessment 

Medical equipment - imaging, non-invasive, hearing aids, wearable devices

Autonomous cars - engine control, braking system, situation assessment, voice activation, machine vision

Industrial machineries - tool fatigue monitoring, motion controls, error reporting and machine vision

Robots - motion manipulation for tasks that cause fatigues, are dangerous and require precision

Smart Agriculture - using data from soil monitoring, plant disease, nutrients etc. and determination of appropriate intervention taking weather conditions and crop requirements into account to maximize crop yields with minimum farm inputs

Retail - real time determination of consumption, inventory and supply of products and an intelligent guess of the future demand through use of models

Power Grids - real time determination of consumption and supply of power and use of intelligent algorithm to draw power and distribute as per the estimated demand

Market forecast for embedded systems

Market assessment for embedded systems have been done by a few market research organizations that point to a tremendous anticipated growth.

· As per QY Research, the embedded system market was US $ 68.9 billion worth in 2017, and is estimated to grow to US $ 105.7 billion by 2025.

· As per B2B forecast, the global embedded systems market is estimated to be US $ 116.2 billion by 2025.

Conclusion

Embedded systems today form part of practically every industry, ranging from consumer electronics to industrial control, smart agriculture, automobiles, healthcare and medical systems, telecom, aerospace and defense, smart grid energy systems, retail, supply chain, smart sensors and devices. Their growth has been triggered by phenomenal increase in processing power coupled with decrease in cost of microprocessors and memory and ability to design low-cost SOC.

The future of embedded systems indeed lies in smart use of artificial intelligent algorithms, ubiquitous computing, edge computing and organic computing for diverse application environments. The market is forecast to be growing multifold. It is aptly described as an industry of the future.

References i. Embedded systems trends and technologies, Dick Slansky, Sept 25, 2019, https://www.arcweb.com/blog/embedded-systems-trends-technologies-0 ii. Future of Embedded Systems and Career options, https://www.digit.in/technology-guides/fasttrack-to-embedded-systems/future-of-embedded -systems-and-career-options.html iii. How Embedded Systems are transforming the Future, Brian Jensen, Sept 27, 2012, https://www.business2community.com/tech-gadgets/how -embedded-systems-are-transforming-the-future-0293881 iv. The Future of AI and the Embedded Systems -Total Phase Blog, August 16, 2017 https://www.totalphase.com/blog/2017/08/the-future-of-ai-and-the embedded-system/ v. How to integrate AI with Embedded Systems, Kamalika Some, Aug 31,2020, https://www.analyticsinsight.net/how-to-integrate-artificial-learning-with-embedded-systems/ vi. Embedded Systems with Artificial Intelligence, Christoph Wagner, Feb 22, 2019, https://www.embeddedcomputing.com/technology/ai-machine-learning/embedded-systems-with-artificial-intelligence