Picture an industrial robot and you’ll probably imagine one of two extremes. Either the fixed arms that have been carrying out repetitive tasks on factory production lines for decades, or the humanoid robots increasingly filling technology company demonstrations. Thomas Tang, co-founder and CEO of Anyware Robotics, believes the future lies somewhere between those two approaches.
Why industrial robotics needs a new approach
“We are building general-purpose robotic labor for industrial tasks,” he says. “We want to solve the global labor shortage issue.”
The physical demands of many industrial jobs are making them increasingly difficult to fill, he argues. “They’re so harsh, so heavy duty that, while the older generation is still willing to do them, the new generation doesn’t want to do them anymore,” he says.
For Tang, the answer isn’t another generation of highly specialized industrial robots of the kind that has transformed automotive production since they were first introduced in the 1960s. He argues that, despite six decades of deployment, industrial robots have had only a limited impact, estimating that around four million have been installed globally over that period, against a global workforce measured in billions.
“If we follow those traditional paradigms, we’d need to take another 500 years to make a reasonable impact,” he says. “That’s why we need a new paradigm.”
Pixmo: pick and move
Credits: Anyware Robotics
Anyware’s answer is Pixmo, a robot that sits between the fixed industrial arms of traditional automation and the humanoid robots attracting so much attention today. Instead of trying to replicate a human form, Pixmo combines an autonomous mobile robot (AMR) base with a collaborative arm designed to work safely alongside people, a suite of sensors, and AnywareOS, the company’s AI software platform. Together, these components form a system designed to tackle physically demanding logistics and manufacturing tasks.
The name itself reflects the robot’s core functions: “pick” and “move.” Pixmo is designed to navigate industrial environments, manipulate objects and adapt to different workflows. Its first commercial application is unloading trailers, but Tang sees that as only the beginning.
Why trailer unloading?
If Pixmo is designed as a general-purpose industrial robot, why begin with unloading trailers?
Trailer unloading is one of the least desirable jobs in warehousing. Workers spend hours inside enclosed metal trailers that can become oppressively hot in summer, repeatedly lifting and stacking heavy boxes. Combined with persistent labor shortages, it has become an urgent problem that many logistics companies have been trying to automate for years.
Tang insists Anyware Robotics is “not a warehouse automation company.” He says trailer unloading was the right place to begin because customers were already willing to invest in solving a costly problem. But commercial demand was only part of the attraction.
The data challenge
For Tang, trailer unloading also addresses one of the biggest constraints facing physical AI systems: data. Unlike software AI, which can be trained on abundant digital datasets, robots must learn by interacting with the physical world. Every warehouse, every damaged carton and every awkwardly stacked pallet presents a slightly different problem to solve. Capgemini Research Institute identifies data scarcity as a limit on physical AI, noting that models ultimately require high-fidelity, real-world data to handle unpredictable physical edge cases.
Thomas Tang: Credits; Anyware Robotics
Tang argues that collecting enough real-world experience remains the limiting factor. When Pixmo encounters situations it cannot confidently resolve, human operators can intervene remotely, helping complete the task while generating new training data. Over time, those interventions become less frequent as similar scenarios are incorporated into the system’s knowledge, creating a feedback loop that steadily improves the robot’s capabilities.
Building the whole stack
Tang believes solving that challenge requires control of the entire robotics stack rather than treating AI as software that can simply be layered onto existing machines. Instead of designing bespoke hardware for every component, Anyware combines proven commercial technologies where appropriate with its own intelligence layer. Tang likens the trade-off to buying a watch. Customers don’t want a different watch for every feature; they want one that strikes the right balance between battery life, usability, reliability and comfort. He argues industrial robots are no different, requiring hardware and software to be designed as a single system that balances performance, reliability, maintainability and cost.
One surprise during early deployments was how positively workers responded. Tang expected some skepticism about robots entering the workplace but found that employees were often relieved to hand over one of the most physically demanding jobs in the warehouse.
Pixmo is designed to work in environments where people are present, using force sensing and autonomous navigation to react safely if someone unexpectedly moves into its path. Tang argues the goal is not to replace human workers but to remove the dull, dirty and dangerous tasks that make recruitment increasingly difficult.
The vision: Anyware everywhere
Trailer unloading is only the starting point. Tang sees the same technology expanding into progressively broader industrial applications, moving from the harshest logistics tasks into manufacturing and other physical industries as the robots accumulate experience.
Tang’s vision is that general-purpose robots can work almost anywhere, adapting to different environments rather than being confined to a single repetitive task. If Tang’s vision succeeds, the next generation of industrial robots will be defined by their ability to keep learning wherever they are deployed.
