A commentary by Andreas Parr, Bob Scannell, Sarven Ipek, Analog Devices.

From the invention of the steam engine during the first industrial revolution to the development of assembly line manufacturing during the second, the world has made great strides thanks to the introduction of new technologies. Many analysts agree that the next industrial revolution has already begun, driven by the growth of Industry 4.0 (Smart Factory) and autonomous systems. The quest for more efficient use of materials and labor that characterizes this new age of industrial discovery requires that the underlying technology continues to evolve at a rapid pace.

Automated and autonomous robots, vehicles and drones that are even more closely integrated into manufacturing, mining, agriculture and logistics processes are key pillars of this relentless industrial revolution.

Sensors as enabling technology

In order for autonomous applications to achieve the level of system performance they require, the equipment used must be able to perceive its environment and find its way around it. This is possible with the help of various types of sensors whose output signals are combined and interpreted, whether using traditional algorithms or those based on artificial intelligence (AI) or machine learning. Reliability and machine availability are the two main challenges in this context, which call for the simultaneous implementation of different sensor technologies in order to achieve the ultimate goal of greater safety and efficiency, lower costs and increased flexibility.

Overview of sensor types for autonomous systems.

© ADI

Autonomous systems rely heavily on high fidelity data captured by fused sensor modalities and fed to AI and algorithms. Radar, lidar, image, ultrasonic and inertial sensors are the most widely used sensors in the industry. The following table provides an overview of the advantages and limitations of each sensor type and emphasizes the need to equip systems with multiple sensor modalities.

Perception through sensors - machines learn to see

The challenges of Industry 4.0 are manifold. Confined spaces and autonomous machines (robots, cobots, etc.) working together in harsh environments, for example, require the use of radar technology that takes up little space, is precise and offers the ability to measure close targets. Mapping and classifying surrounding areas is essential for efficiency, productivity and safety.

Radar technology in autonomous machines

Driven by recent advances in RF transceiver ICs, radar is rapidly becoming one of the most important sensor technologies for sensing applications. One example of this is the fully integrated and purely digital transceiver MMICs operating at 77 GHz. High-frequency and highly linear FMCW chips in conjunction with high output power, low-noise transmit and receive channels and MIMO antenna arrays now make it possible to realize very powerful and high-resolution radar systems at acceptable costs. Radar-based digital beamforming allows the detection of the radial velocity, angle and distance of multiple targets even under the most adverse environmental conditions, and this is crucial for the safe and efficient interaction of robots, cobots and driverless transportation vehicles in dynamic environments.

Precision through lidar systems

Autonomous systems in the industrial sector often have the task of localizing and picking up objects instead of driving around them. The powerful object detection and high classification accuracy of lidar technology provides the necessary precision for these common tasks.

Because they operate at frequencies in the terahertz range, lidar systems achieve a very high angular resolution, which in turn results in high-resolution depth maps. The latter enable lidar systems to classify objects for merging with image, IMU and radar information so that reliable mission-critical decisions can be made. Lidar systems are designed for use in dynamic conditions, for example in the open air and in bright sunlight. By using short, powerful pulses with wavelengths of 9xx nm and 15xx nm, lidar technology achieves greater visibility under these demanding conditions. Not least, the short pulses enable a higher depth resolution for detecting multiple targets within a pixel, while the 9xx nm and 15xx nm wavelength infrared light used is less affected by sunlight.

However, a number of challenges still need to be overcome to pave the way for the mass use of lidar systems. These include the complex and costly signal chains, problems with the optical design and testing and calibrating the systems. In this regard, development projects are currently underway with the aim of integrating the signal chains and reducing their complexity, dimensions and power consumption, as well as lowering overall costs.

Navigation sensor technology - machines learn to feel

The more sensors are used on industrial machines and the greater the amount of data they provide, the more important their location and movement relative to each other becomes. Autonomy is often linked to mobility, which is why it is crucial to determine the exact position of a vehicle, guide the movement of a machine and precisely control its instruments. Detecting such movements with high precision allows such applications to be used in more difficult, useful applications where safety and reliability are also important. In the field of smart farming, for example, the aim is to continuously increase the efficiency of harvest management, and the centimeter-precise positioning of instruments is the primary driving force in the effort to produce as much as possible with as little effort as possible.