Have you tried summoning your new car out of a parking spot using your smartphone? Did you know that many delicate surgical procedures nowadays are handled by robotic hands? Have you heard about drones that autonomously track GPS flight routes and land themselves safe and sound?
These days, robots have even become part of our vacuum cleaners and education or informational electronics.
Astonishing advances in semiconductor technology have resurrected autonomous machines and artificial intelligence yet again; but this time for real! Machines are increasingly performing tasks that were previously reserved for humans, and the transformational impact on our world is fascinating – a discussion reserved for another platform.
The rise of intelligent machines, specifically drones, robots, and semi-autonomous vehicles, has evolved from novelty to a myriad of commercial and industrial applications. The electronic content of these machines is growing fast, which enables new capabilities such as precision control and ambient awareness, which is a general awareness of one’s surroundings without physical proximity or specifically requesting such information. The older generations of industrial robots or platooned vehicles were mostly pre-programmed to perform repetitive tasks with high accuracy. Modern autonomous machines, on the other hand, have distinctive characteristics including interaction with humans, dealing with unplanned events, and rapid-learning. These are all features that demand a higher level of ambient awareness, intelligence and control.
Two applications specific to artificial intelligence include monitoring and control. Monitoring applications typically deal with big data generated by the Internet or a large number of sensors. Data mining and pattern recognition of the gigantic and ever-growing amount of data requires extremely high processing power of servers and processors. A large number of applications, like security, medical diagnostics, and marketing to name a few, rely on data analytics and deep learning techniques. Control applications, on the other hand, require real-time analysis of sensor data and autonomous control of actuators and motors. Autonomous vehicles, drones and the new generation of robots fall into this category. Discussing control applications is where I will spend the rest of my time in this article.
I believe a confluence of three enabling technologies has contributed to the resurgence of intelligent machines:
1) Sensing: Thanks to advances in semiconductor and microelectromechanical systems (MEMS) technologies, a wide range of sensors for ambient awareness and vision are affordably being deployed in autonomous systems. While vision sensors such as cameras, radar and LIDAR (sometimes considered the acronym of “light detection and ranging”) are critical for many autonomous systems, other ambient sensors which require significantly less processing power are more common. High precision torque, temperature and magnetic sensor fusion can provide abundant information about tactile, proximity and ambience; which enable robots to interact with humans and handle unpredictable situations both safely and effectively. This can be compared to the many biological systems in humans and animals that react on stiffness, heat, magnetic fields and many other non-vision sensing capabilities.
Building upon this sensing technology, vision analytics is working itself into the mainstream, as algorithms and processors are becoming more powerful and affordable. Machine vision has long been used in both industrial and consumer applications (think gaming consoles), but the mission-criticality of autonomous systems demands for higher-performance and higher-reliability machine vision. Because of this, we are now seeing high resolution cameras abundantly being deployed in cars, robots, and drones. These autonomous systems are taking advantage of advances in vision algorithms and stereo cameras with advanced processors for depth detection and fusion.
Fully-integrated complementary metal–oxide–semiconductor (CMOS) radar with the extended range of 250 meters and accurate to a few microns (but not at the same time, mind you) complements camera when optical vision is not available or practical, such as when driving in heavy rain, snow or fog. LIDAR can provide an even more detailed map of a machine’s surroundings in real-time with highly focused laser beams.
2) Processing power: The availability of tremendous processing power combined with advanced neural network algorithms has significantly contributed to efficiency and reliability of machine vision, pattern recognition and machine learning. Extensive offerings of graphics processing units (GPUs), bio-inspired processors, and multicore very long instruction word digital signal processors (VLIW DSPs) have enabled a wide range of vision subsystems. Teraflops super GPUs with extensive auxiliary accelerators for deep learning and image processing are critical for complex advanced driver assistance (ADAS) systems. However, many autonomous systems are battery-powered with limited processing energy budget, demanding energy efficient processing. High performance multi-core DSPs equipped with accelerators for machine learning algorithms can provide the type of energy-efficient processing power required for systems like mobile robots and drones. A typical multicore DSP, such as a TI C7X, can provide processing with lower power consumption, lower bill of materials (BOM) and higher scalability than a GPU.
3) Motor and actuator control: Electric motors have been replacing many hydraulic and mechanical systems in autonomous machines. Improved efficiency of electric motors and intelligent motor drivers has enabled robots and drones to perform high precision motions. While electric cars take advantage of the improved efficiency of induction AC motors and drivers, many light robots and drones use brushless DC for their high efficiency and almost zero maintenance. TI has a wide portfolio of gate drivers, such as the DRV8X product family, that provide smart gate drive functions and high integration for optimizing performance and compact board design. Motors with integrated sensors, fault diagnostics and intelligent power management are essential for high precision and high torque applications in autonomous systems.
Intelligent machines, despite technical and other challenges, are appearing in our daily life. Their transition from novelty to everyday use is slowly becoming more noticeable each day. Rapid advances in electronics and creative applications indicate resurgences in intelligent systems that are here to stay. The new generation of autonomous machines is here to help, let’s embrace it!
Just last evening we were dicussing this very matter. What an old yet new world we are living in. Thank you for a more in depth article about our future and the part we will play in it. Amazing...
I suddenly became interested to AI and i cant stop
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.