As consumers demand higher-quality goods at steadily increasing volumes, the chief benefits of industrial automation — speed, accuracy, and consistency—become more important to businesses worldwide with each passing day. With processes evolving and manufacturing speeds increasing to meet these demands, turning to industrial automation becomes inevitable.

The long-term necessity of deploying automation to combat labor shortages has also become increasingly clear. According to June 2024 data from the U.S. Chamber of Commerce, unemployment in durable goods manufacturing stands at only 2.8%. And a report published in April 2024 by Deloitte and The Manufacturing Institute suggests that more than half of the expected 1.9 million manufacturing jobs needed to be filled by 2033 could go unfilled. 

In the past, the price of deploying automation to augment human labor was sacrificing flexibility. Traditional robots deployed in industrial settings executed strict commands, like applying torque or welds at the same location on identical parts, or grasping and moving the same objects from one precise location to another. The primary beneficiaries of such automation were heavy industries like automotive manufacturing, and the number of appropriate tasks was limited.

Introducing machine vision capabilities to robot systems opened a much wider range of potential automation applications in a larger number of industries. Introducing artificial intelligence (AI) into vision-guided robots allows these systems to take yet another leap forward, with broad potential across many industries.

Deep Learning Boosts Bin Picking

Early machine vision applications relied on rules-based algorithms that enabled systems to detect and recognize specific objects within images. For example, when dimensions of parts, their relative position to the camera, and lighting conditions remained consistent, parts could be picked from moving conveyors. Automated bin picking was problematic in its early years due to issues like partial occlusion from overlapping parts and lighting variations.

Deep learning, a subset of machine learning, has added new layers of ability to vision-guided robotics applications. Whereas earlier machine vision required vast image sets to train algorithms and define object parameters, deep learning allows neural networks to virtually teach themselves with far fewer images to start. This ability, coupled with exponentially increasing processing speeds, has made it possible for new vision systems to detect and identify a much wider range of objects and to better understand their relative positions.   

Manufacturing and Logistics Benefits

Unlike their historical counterparts, vision-guided industrial robots with AI can adjust their own performance. For example, robotic welders and adhesive dispensers not only become more accurate with the addition of AI-enabled vision guidance, but they can also inspect the quality of their welds or beading and make corrections to ensure that quality specifications are met.

Vision-based industrial inspection applications no longer depend on cameras mounted in fixed positions. Robot arms equipped with vision systems and AI-enabled guidance can maneuver cameras around objects to get the best positions for visual inspection and can adapt to varied object placements relative to the camera.

Smart bin picking combines 3D vision with quick and accurate robot arms and software. Compared to traditional pick-and-place robot applications, smart applications can safely handle, at high speeds, smaller objects, a greater variety of object shapes, a wider range of light reflection, and, with the appropriate end-of-arm-tooling, objects constructed from more delicate materials.

Vision-guided systems leveraging AI can also tackle random bin-picking applications — where the contents of each bin might vary widely from the contents of the previous bin. The same technology can be used with conveyor systems. Removing the need to sort items before placing them on a conveyor or in a bin increases efficiency.

Random palletizing and depalletizing applications can also leverage AI and vision guidance. Vision-guided palletizing robots can accurately detect and grasp non-homogenous package types while also using vision to detect the amount of available space on a pallet and to determine optimal placement strategies. Depalletizing robots can respond nimbly to changes in package dimensions while unloading pallets.

Conquer Food and Beverage Challenges

In food and beverage, AI can help vision-guided robot systems handle high-variability environments and go beyond just making an accurate pick-and-place. AI-enabled vision systems can bring humanlike inspection and decision making to the process. Consider the case of apples, which can vary widely in size and shape, so determining total volume through visual inspection is challenging. Bruising or disease on fruit can also present in a wide variety of shapes and colorations. 

Likewise, cuts of poultry or other proteins are not identical, and contamination presents itself visually in different ways. AI-enabled vision systems can reliably, even at high speeds, analyze images of produce or proteins and send accurate movement commands to robot arms, which then can select or package only products that meet safety requirements. In these applications, the accuracy of AI-enabled vision is paramount. False negatives lead to robots disposing of perfectly good products and lowering yield. False positives could result in illness or death, with financial and legal ramifications.

While optimized illumination setups will always be necessary for accurate vision system performance, AI-enabled vision better tolerates variances in lighting, such as reflection and absorption. This makes it possible to accurately inspect products like produce or meat wrapped in cellophane, to check liquid fill levels, and to identify the presence of contaminants in plastic bottles, improving both product safety and quality.

Vision-Guided Robots on the Move

The combination of vision systems and robotics had enabled the creation of an entirely new technology. Many autonomous mobile robots (AMRs) leverage AI-enabled vision and fast processing speeds for functions like simultaneous localization and mapping (SLAM), which allows an AMR to plot its position as it moves within a facility, and collision avoidance, which allows an AMR to operate in close proximity to human workers.

AMRs sometimes see intra-plant deployment. In semiconductor plants, AMRs may transport component reels in a chip fab between storage racks on the shop floor and machines that require reloading. AMRs might transport parts or products between assembly cells. Huge AMRs even transport engine blocks within tractor assembly plants. 

The logistics and warehousing sectors arguably make the widest use of AMR technology. Common applications include transporting component racks on casters or retrieving and moving totes, sometimes to or from bin-picking stations. When equipped with robot arms for pick-and-place, vision-guided AMRs can also operate in high-volume fulfillment centers, where AI again allows the system to handle high part or product variability. 

While still under development and with limited commercial applications, tracked agricultural AMRs equipped with vision-guided armatures can navigate down rows of crops and use vision to detect fruits or vegetables that are ripe for picking. And thanks to AI-enabled vision, autonomous quadruped robots, once purely a thing of science fiction, have a growing list of live remote inspection applications in industrial and hazardous environments.

Deployment Opportunities and Challenges

Vision-guided robotics and AI enable a wide range of applications across many different industries, which suggests the almost limitless potential for the technology. New applications include retail environments, where few applications currently exist. And vision-guided bipedal robots may offer potential future value in applications such as elder care and companionship. 

Companies beginning with new automation deployments can take advantage of modern robots that come standard with vision-based guidance and may even feature AI-enabled software right out of the box. To support robotic deployments, greenfield facility design should include floor layouts with adequate space for armatures or AMRs and appropriate lighting systems to support vision.

Upgrading older robots with machine vision or incorporating the technology into brownfield facilities presents more of a challenge. Equipment interoperability will be a primary concern. New lighting may be required. Floor layouts may require adjustment. 

Now that vision-guided robotics has advanced far beyond mere pick-and-place applications, manufacturing, logistics, food and beverage, and other companies may not wish to risk falling behind by ignoring the potential opportunities to deploy the technology. After all, there are so many options.