Welcome to the Era of 'Physical AI' in Industrial Robotics

2026 is a year that will surely be marked by a new way of looking at robotics within industry. The redefinition of industrial robotics began in 2025 with two events that definitively marked a before and after in industrial robotics.

The first event, ISO 10218-1 and ISO 10218-2, was updated in February 2025, where the philosophy of industrial safety, which previously focused on the physical characteristics of the "cobot" robot (force sensors and rounded edges), now focuses on the application, the environment, and the task. The workspace must be collaborative (robot, end effector, workpiece, and human presence), and the collaborative workspace (CWS) must consider spatial limits, modes of operation, and integrated risk assessment. This will allow any robot to be collaborative by installing laser scanners and photoelectric barriers to reduce the speed of the robot when the operator approaches, creating a collaborative workspace even with conventional industrial robots. Risk assessment must be realistic and transparent, forcing companies not to assume that cobots do not require protection. The focus is on protecting the operator from a specific risk, as this update to ISO 10218:2025 reinforces that safety is not part of the hardware but is the result of the design of a complete system.

In October, the second major event was announced, with Japanese conglomerate SoftBank Group reaching an agreement to acquire ABB's robotics division for approximately US$5.4 billion. The Swiss-Swedish engineering giant had already announced its intention to split off its robotics division, but this injection of capital will allow it to strengthen its electrification and automation business units, while SoftBank will accelerate the development of more intelligent, autonomous robots capable of performing complex tasks in the real world by merging artificial intelligence with robotics, a concept that gives rise to "Physical AI."

Following the merger of Softbank and ABB, the introduction of Physical AI in industry promises a radical transformation that goes beyond automating repetitive tasks:

1. Programming for intelligent adaptation

• Robots that learn and adapt: Through advanced computer vision and reinforcement learning, robots will be able to adapt in real time to variations in their environment.
• Complex manipulation and fine assembly: Robots sense and understand what they are doing, applying force and precision without the need for meticulous programming of each process. 

2. Safer and more intuitive human-robot collaboration (cobots)

• True co-workers, cobots will reach a new level where robots will be able to predict the intentions of the operators around them, generating fluid and safe movements to avoid collisions and adapting to the operator's work pace with a contextual understanding of the safe space.

3 Optimization and predictive maintenance at the corporate level

• Collective intelligence, through a central AI that creates a learning network where information can be shared and deployed instantly with all robots in the corporation, locally or globally.
• Self-diagnosis and repair: Robots will be able to predict failures by analyzing their own data, reducing downtime due to failures or maintenance.

Physical AI is taking humanoid robotics and service one step further with projects such as the second-generation Optimus robot (TESLA), a second-generation humanoid that seeks to learn to navigate and manipulate objects autonomously; and Boston Dynamics (Hyundai) with its Atlas and Spot robots, which are not far behind, focusing on "integrated manufacturing intelligence." By 2026, they aim for their robots to understand and execute complex tasks autonomously, complementing their agility of movement. For its part, Physical AI is developing humanoids that understand verbal instructions and translate them into physical actions. To ensure that these developments are not limited to a select few, Unitree Robotics is democratizing access to Physical AI with its G1 robot, offering low-cost platforms for developers.

It is not known if industrial robotics will catch up with humanoid and service robotics, but what is certain is that Physical AI will transform industrial robotics from a programmed tool to an intelligent and adaptable collaborator in a collaborative and safe workspace, which will be reflected in optimal and high-quality production processes.