> **来源：[研报客](https://pc.yanbaoke.cn)** # Summary of the Document: Physical AI - The Moment of Acceleration ## Core Content Physical AI (PAI) refers to the integration of artificial intelligence with physical systems such as robots, autonomous vehicles, drones, and smart manufacturing systems. It enables machines to perceive their environment through sensors, make real-time decisions, and act with tangible consequences. The document outlines the current state and future trajectory of PAI, emphasizing its role in industrial robotics and the broader business transformation it can enable. ## Main Points - **PAI is transitioning from science fiction to mainstream business use**, with 2025 marking a significant inflection point. - **Hardware advancements** have made PAI more viable, including cost reductions and improved capabilities in sensors and edge computing. - **Software innovation**, particularly in simulation and open ecosystems, has enabled AI models to be trained in digital twins and transfer effectively to the real world. - **Global governance and investment** are accelerating PAI adoption, with regulatory frameworks and national strategies playing a crucial role. - **PAI is already delivering value in various sectors**, including manufacturing, logistics, healthcare, and transportation. ## Key Information ### PAI in the Real World - **Current impact**: Only 5% of firms report PAI transforming their organisation today, but this is expected to rise to 41% within three years. - **Applications**: - **Create**: Robotic arms in car manufacturing. - **Observe**: Drones scanning infrastructure. - **Handle**: Humanoid robots moving totes in warehouses. - **Protect**: Drones as first responders. - **Transport**: Autonomous vehicles delivering people or goods. - **Interact**: Humanoid robots in customer service and care environments. ### Where PAI Scales Value: Industrial Robotics - **Proving ground**: Industrial robotics is where PAI is demonstrating tangible ROI and operational improvements. - **Three layers of value**: 1. **Operational value**: Enhances factory efficiency, reduces downtime, and enables complex automation. 2. **Value spillover**: Integrates AI across the value chain, improving supply chain agility and reducing waste. 3. **Disruptive innovation**: Enables new business models such as Factory-as-a-Service (FaaS) and Operation-as-a-Service (OaaS). ### The PAI Technology Stack PAI is not a single technology but a complex stack with six interdependent layers: 1. **Platform**: Digital twin environments for AI training and simulation. 2. **AI and Cognition**: The "brain" responsible for decision-making and adaptive interaction. 3. **Sensing and Perception**: The "eyes, ears, and skin" of the system, using cameras, sensors, and detectors. 4. **Mechanical Structure**: Defines the robot's physical form and capabilities. 5. **Compute and Control**: Edge computing connects AI to physical execution. 6. **Power and Actuation**: Ensures energy efficiency and robustness. The stack mirrors a biological system, where each layer contributes to the overall intelligence and adaptability of the physical system. ### Getting Past the Bottlenecks - **Technological bottlenecks** include the simulation-to-reality gap, lack of physical common sense in AI, and high costs of sensors and edge computing. These are largely beyond the control of most organisations. - **Operational bottlenecks** involve internal readiness, such as automation foundation, digital infrastructure, and workforce skills. These can be addressed through strategic leadership and organisational change. ### Realising the Value: The Dual Maturity Lens PAI delivers value only when both **technology application maturity** and **operational maturity** are aligned. The document outlines four stages of technology application maturity: 1. **Stage 1 - Automation**: Machines perform predefined tasks with precision. Prerequisites include PLCs, HMIs, and safety systems. 2. **Stage 2 - Collaborative Digitalisation**: Machines become aware of their environment and work alongside humans. Prerequisites include cobots, vision systems, and IIoT gateways. 3. **Stage 3 - Digital Twin**: Virtual and physical systems are synchronized. Prerequisites include digital twin platforms, physics engines, and integration with MES and ERP. 4. **Stage 4 - Physical AI**: Machines perceive, reason, and act autonomously. Prerequisites include advanced AI models, VLA fine-tuning, and synthetic data. ## Conclusion The adoption of PAI is accelerating due to technological and governance convergence. While the technology is becoming more accessible and capable, its successful implementation depends on both technological readiness and organisational preparedness. Business leaders must understand the dual maturity lens and strategically invest in the PAI technology stack to fully realise its potential.