14 August 2026 | Interaction | By Editor Robotics Business NEWS <editor@rbnpress.com>
As Physical AI moves from demonstrations to real-world industrial applications, autonomous robots are creating new possibilities for safer, continuous and intelligent inspection. In this interview with Robotics Business News, Arthur Chung, Vice President, Sales & Supplier Management, Avnet Asia, discusses Avnet’s collaboration with Weston Robot, the role of edge AI and 3D LiDAR SLAM, and how autonomous inspection can improve safety, reliability and operational efficiency across industries.
What were the key industry challenges that led Avnet and Weston Robot to develop this autonomous Physical AI inspection platform?
Industrial inspection has traditionally relied on manual patrols, scheduled checks and people identifying problems when they occur. That can work well in smaller or more controlled environments, but it becomes much harder as facilities get larger, more complex and more automated. Many industrial sites operate around the clock, with large areas to cover, extensive equipment and environments that may not always be safe or practical for people to access. Relying only on periodic inspections can also mean there are gaps between one inspection and the next. If something changes during that period, it may not be picked up until the next scheduled check.
At the same time, AI is moving beyond software applications and into machines that can sense and respond to what is happening around them. That is where Physical AI has real potential for industry. Instead of AI simply analysing information on a screen, it can be part of a system that observes a physical environment and responds to what it finds.
We saw an opportunity to apply that idea to industrial inspection through our work with Weston Robot. Weston brings strong expertise in robotics and autonomous navigation, while Avnet brings experience in embedded computing, edge AI and bringing different technologies together into a working solution. The aim was not simply to put a robot into a factory and have it move around, but to create a platform that could inspect an environment continuously, identify conditions that may need attention and give operators useful information sooner.
Ultimately, the challenge we were trying to address is quite straightforward. We want to be able to provide better visibility of what is happening across a complex industrial environment, without relying entirely on manual inspection.
How does the robot combine AI perception, edge computing and autonomous navigation to make real-time inspection decisions in complex industrial environments?
The platform brings together three key elements of perception, intelligence and mobility.
The perception side comes from the different sensors on the robot, including visual, thermal and spatial sensing—each providing a different view of the environment. A camera might identify a visible defect, for example, while thermal sensing can highlight an unusual temperature pattern that would not necessarily be apparent to the human eye.
The AI and edge computing provide the intelligence. Rather than collecting all the data and sending it elsewhere to be analysed, the platform can process information on the robot itself. This allows it to identify anomalies, recognise changes in conditions and flag information that may need attention while the robot is carrying out its inspection. Then there is autonomous navigation. The robot needs to get to the right places in the first place, and it needs to do that safely. It can understand its surroundings, plan its route and adjust its movement when it encounters obstacles or changes in the environment.
What is important is how these technologies work together—a sensor can collect information and a navigation system can move a robot, but bringing sensing, computing and mobility together allows the system to do much more than any one capability can do on its own.
The robot can observe what is happening, process that information and take an appropriate action. That combination of sensing, intelligence and physical action is what makes this a good example of Physical AI.
What advantages does processing AI workloads directly at the edge provide for industrial inspection compared with cloud-based approaches?
For autonomous inspection, there are situations where the system needs to respond quickly. If a robot detects a potential equipment problem or safety issue, there can be value in processing that information immediately rather than sending all the data to the cloud first. In these situations, AI inference can take place directly on the robot with the AMD Ryzen™ AI Embedded processors. This allows the system to process information locally and respond with low latency.
There is another practical consideration—connectivity. Industrial facilities are not always covered by a strong, consistent network, particularly in remote areas, large sites or locations where there may be interference. An autonomous system should not become ineffective simply because its connection to the cloud is temporarily unavailable.
There is also a data consideration. A robot can generate a significant amount of information from its cameras and other sensors. Much of that raw data may not need to be sent back to the cloud. Processing it locally means the system can identify what matters and transmit the relevant information, such as an anomaly or an alert.
All these, however, does not mean that everything has to happen at the edge. There is still a role for the cloud when it comes to storing information, analysing trends across a site or fleet, and providing broader operational visibility.
The two approaches can complement each other. The edge deals with decisions that need to happen close to the point of action, while the cloud can provide the bigger picture.
How does the robot use 3D LiDAR SLAM to navigate reliably in GPS-denied environments such as factories, tunnels, ports and utilities?
GPS is not always practical for industrial robotics. Inside factories and tunnels, for example, satellite signals may not be available. Even in large outdoor sites such as ports and utility facilities, GPS may not provide the level of positioning accuracy required for autonomous movement.
Our platform uses 3D LiDAR-based Simultaneous Localization and Mapping, or SLAM. LiDAR continuously scans the surroundings, allowing the robot to build a three-dimensional map of the environment, while SLAM helps it determine where it is within that map. This allows the robot to use the environment around it as a reference rather than relying entirely on GPS.
That becomes particularly useful in industrial environments because they are not always static. For instance, vehicles move, equipment can be relocated, people come and go, and routes may change. The robot therefore needs to understand its surroundings as it moves rather than simply follow a route that was programmed in advance.
For an inspection platform, navigation is fundamental. It does not matter how good the AI analysis is if the robot cannot reliably get to the area that needs to be inspected.
3D LiDAR SLAM provides the underlying navigation capability that allows the robot to operate across more complex environments, including locations where GPS is unavailable or unreliable.
What types of anomalies can the platform detect, and how can early detection improve industrial safety and operational efficiency?
The platform can monitor a range of conditions, including thermal and visual abnormalities, equipment conditions, personnel and vehicle activity, PPE compliance and unauthorised access.
For equipment, thermal imaging can pick up unusual temperature patterns that may indicate overheating, equipment stress or an emerging fault. Visual inspection can identify things such as leaks, physical damage, wear or corrosion. The same sensing capabilities can also provide information about what is happening around the site, such as the presence of people or vehicles in particular areas or whether safety requirements are being followed.
What matters is that it is not just about detecting an anomaly, but detecting it early enough for someone to do something about it. Equipment problems and safety incidents often have warning signs, such as a gradual increase in temperature, a component beginning to deteriorate or a change in operating conditions. Those things can be difficult to spot consistently when inspections are carried out only at set intervals. An autonomous inspection system provides another layer of monitoring. It can return to the same areas regularly and look for changes over time, giving operators and maintenance teams more information to work with.
This does not replace people or their expertise, but rather, provides them better information and can help them focus their attention where it is needed. That can support earlier intervention, improve safety and help maintenance teams address potential problems before they turn into more costly failures or unplanned downtime.
How are you addressing the challenges of deploying autonomous robots across different industrial environments, facility layouts and operational requirements?
There is no single standard, industrial environment. A factory, port, utility facility and logistics centre all have different layouts, operating conditions and safety requirements. Even two facilities in the same industry can have very different needs. That is why deployment is about much more than choosing a robot. The computing, sensors, AI, navigation, connectivity and software all need to work together, and they have to make sense for the environment in which the system will actually operate. This is where things can become complicated, since a solution that works well in a controlled demonstration may encounter very different conditions in a working facility. There may be moving equipment, changing routes, restricted areas, different lighting conditions and existing systems that the new solution needs to work alongside.
Avnet brings experience across these technology areas, together with relationships across a broad ecosystem of technology providers. That gives us a system-level view of the problem. We can look at the computing requirements alongside the sensing, AI workload, connectivity and other requirements, rather than considering each piece separately. In the work with Weston Robot, for example, its expertise in robotics and autonomous navigation is complemented by Avnet's expertise in advanced computing and embedded technologies.
The important thing is that all of these technologies need to work together reliably. A system is only useful if it can operate in the environment where the customer needs it.
That is why engineering and integration are such an important part of bringing autonomous systems into industrial use.
What role does Avnet's technology ecosystem and engineering expertise play in helping companies bring Physical AI into real-world industrial deployment?
One of the interesting things about Physical AI is that no single technology makes it happen. In fact, you need computing, AI, sensors, connectivity, robotics and software, and those technologies have to work together. This creates a different kind of challenge for customers as they may have strong expertise only in their particular industry or in one part of the technology stack, but bringing everything together can require knowledge across several different disciplines.
This is where Avnet comes in to help. Our experienced engineers work with technology providers across these areas and have engineering expertise spanning different parts of the system. We can help customers look at the available technologies, understand the trade-offs and work out how the different pieces need to fit together.
The Weston Robot collaboration is a good example. Weston brings its expertise in autonomous robotics and navigation, while Avnet brings advanced computing, embedded AI expertise and access to a wider range of technology partners.
For us, the value is in helping connect those pieces. Physical AI is still a developing field, and customers do not necessarily want to spend their time figuring out every part of the technology stack themselves. The bigger industry challenge is therefore not simply developing more powerful AI. It is making that AI work reliably in physical environments, with all the constraints that come with them.
We believe that companies that can bring the right technologies together and solve that integration challenge will play an important role in taking Physical AI from individual demonstrations to practical industrial applications.
Looking ahead, which industries and applications will see the greatest impact from Physical AI-powered autonomous inspection, and how do you expect the technology to evolve over the next three to five years?
We see strong potential in industries where safety, asset reliability and continuous monitoring matter. Manufacturing, energy, transportation, logistics, utilities and critical infrastructure are all areas where autonomous inspection could make a difference. These environments tend to be complex and operate for long periods, often with a lot of equipment and activity taking place at the same time. Missing a relatively small problem can have a significant impact, whether that means equipment downtime, a maintenance issue or a safety concern.
Autonomous inspection can provide another way of monitoring these environments. Rather than relying entirely on people to make periodic checks, organisations can use robots to collect information more frequently and consistently, particularly in areas that are difficult, remote or hazardous for people to access.
Over the next three to five years, I expect the technology to become considerably more capable. Improvements in edge AI, sensors and computing will allow robots to process more information locally and recognise more complex situations. I also expect autonomous robots to become less of a standalone technology. They will increasingly connect with other machines, operational systems and people, becoming part of a much broader flow of information across a facility.
That is where I think Physical AI becomes particularly interesting—it is not simply about putting AI into a robot but about giving machines the ability to understand what is happening in the physical environment and act on that information. As that develops, we could see a shift from robots being used primarily to collect information towards systems that can contribute more actively to how industrial environments are monitored, maintained and managed.
For businesses, the potential is significant—safer operations, earlier identification of problems, better use of resources and greater visibility across increasingly complex physical environments.
