mmWave, 3D object detection at 30m distance possible
Implementing detection of transparent objects such as glass and plastic
mmWave sensors combined with deep learning for autonomous decision making In
a previous article, we covered how TI’s millimeter wave (mmWave) sensors bring intelligence to the edge for robotic arms in factories. In this article, we’ll look at how mmWave technology is bringing intelligence to the edge for autonomous robots.
Through intelligence, sensors can make real-time decisions, such as slowing down or stopping a robot's movements, enabling reliable operation of the robot in industrial robotics applications.
TI mmWave sensors are used in systems designed to avoid collisions in industrial robots, solving key challenges associated with using robots to interact with people and other objects.
When additional machine learning processing is needed at the edge, mmWave sensors can work seamlessly with industrial-grade processors, such as Sitara processors, to provide additional intelligence.
▲ Figure 1: Various unmanned transport vehicles
Advanced driver assistance systems for automobiles(ADAS) uses mmWave to perform surround view monitoring and obstacle detection. Similarly, mmWave sensors solve similar problems in autonomous guided vehicles (AGVs), such as logistics robots, delivery vehicles, forklifts, and pallet jacks, as shown in Figure 1.
3D Point Cloud Detection The mmWave sensor has a three-transmitter and four-receiver antenna configuration, enabling 3D object detection up to 30 m away, and provides both azimuth and elevation angle information to detect objects at high heights. This is useful for vehicles such as forklifts where the sensor is positioned high above the ground.
A single sensor can detect objects across a 120-degree field of view, minimizing the number of sensors required for a surround monitoring system.
High resolution for accurate detection mmWave sensors operate at a bandwidth of 4 GHz, allowing them to identify individual objects as close as 4 cm and operate in spaces where they may be obscured by darkness, dust, or physical barriers.
High-resolution capabilities are necessary for the sensor to accurately count and identify objects or people, and trigger appropriate actions in real time, such as stopping a robot before a collision.
In addition to real-time object detection and collision avoidance, mmWave sensors provide additional capabilities that make industrial robots smarter.
Ground speed and edge detection TI mmWave technology provides sub-millimeter accuracy and high resolution for precise ground speed sensing via Doppler shift, enabling sensor systems to calculate a vehicle’s ground speed and detect ground edges, such as docks where wheels may slip, to prevent irreversible situations.
Transparent Object Detection TI mmWave sensors can detect transparent objects like glass and plastic, as well as dark objects that some light-based technologies have difficulty detecting. Increasing detection accuracy can help prevent accidents involving glass obstacles or objects.
SIL-2 compliant TI’s 60-GHz mmWave sensors help systems meet the IEC SIL-2 standard for incident management in tight human-machine interaction. When designing a SIL-2 certified system, using TI mmWave sensors to provide these capabilities eliminates the need to implement additional safety processor systems or redundant sensor systems to achieve certification.
▲ Figure 2: Typical front-end sensor solution and
Capable of performing intelligent functions at the edge
Integrated TI mmWave Sensor Comparison
TI mmWave sensors are unaffected by environmental and lighting conditions like rain, dust, and smoke, and can sense through materials like plastic, eliminating the need for external lenses, apertures, or sensor surfaces to effectively detect objects in a robot’s path.
To give more intelligence to the edge Deep learning, which constitutes machine learning, is increasingly being adopted in the industrial sector. TI provides hardware and software support to make it easier to apply deep learning inference to the edge for a variety of applications, including robotics.
For applications where smart sensing isn’t enough, the Sitara AM57x processor family features Arm Cortex-A15 cores and dual-core C66x processors operating at up to 1.5GHz.
Deep learning inference can be used to address machine learning needs for applications such as making predictions about maintenance and calculating remaining useful life by performing traditional machine vision algorithms as well as making decisions based on sensor inputs beyond existing capabilities.
The Sitara AM57x processor provides dedicated hardware for industrial communication (EtherCAT, PROFINET, TSN, PROFIBUS, EtherNet/IP) and can serve as the central processing unit of a robot controller.
In summary, our integrated mmWave sensing solution, which delivers superior object detection, together with the Sitara processor for enhanced machine learning, provides an intelligent solution for area detection around a robotic arm or collision avoidance in autonomous robots.
To get started developing, you can refer to TI's resources.
This article was written by Prajakta Desai, mmWave Sensor Product Marketing Manager, and Lali Jayatilleke, Applications Engineer, Text Instruments.
