Robotics is undergoing the fastest evolution in its history, driven by the convergence of physical AI, advanced perception, edge computing, and increasingly capable electromechanical systems.
What was once the domain of rigid, pre-programmed industrial machines has transformed into a new generation of robots that can see, understand, plan, and act in dynamic real-world environments.
Breakthroughs in computer vision, depth sensing, Physical AI–driven behaviors, and on-device inference now allow robots to adapt to variability, collaborate safely with humans, and perform tasks that previously required human judgment. As industries face rising quality expectations and increased pressure for 24/7 operations, robotics is shifting from experimental deployments to large-scale, economically viable automation—marking the beginning of an AI-native robotics era.
This is the backdrop for Capgemini’s robotics project leveraging Intel’s end-to-end technology stack and RealSense™ cameras.
Capgemini’s engineering capabilities in robotics
Capgemini leverages decades of expertise in automation, AI, embedded systems, simulation, and edge computing to deliver next-generation robotic solutions. The latest demo, led by Kevin Cloutier, North American Director of Robotics, the rapid acceleration happening across the industry. As Cloutier explains, robotics has seen exponential growth in a new era in which AI and robotics are not separate domains but have fully converged.
Much of his team’s research is already production-ready and designed for real-world scale. In addition, their work on enabling robots to contextualize their environments using Vision-Language-Action (VLA) models on Intel hardware represents a breakthrough shaping the future of robotics. This innovation is driving forward the next generation of intelligent, adaptable machines.
Known as “Project REACH,” the first phase of the initiative focuses on four core pillars:
- Human-level perception & computer vision
Leveraging RealSense™ depth cameras and Geti™, Capgemini’s robotics platform can detect and classify objects, track moving targets, interpret orientation, and adapt dynamically to changing environments. - Robotic manipulation and motion control
Using collaborative robots like the UR5e, the system integrates advanced capabilities such as inverse kinematics, dynamic path planning, grasping pipelines, and ROS 2–based control for precise and flexible operations. - Modular, interchangeable architecture
Designed for adaptability, the platform’s modular architecture allows teams to seamlessly swap sensors, cameras, and robotic components to meet evolving requirements. - End-to-end AI development pipeline
Engineers can train, optimize, and deploy perception models locally using Geti and OpenVINO™, reducing development cycles from weeks to hours.
The second phase of innovation focuses on giving robots true environmental awareness. This means moving beyond simple object recognition to more deeply contextualizing the environments in which they operate. By integrating Vision-Language-Action (VLA) models, robots can interpret visual cues, understand natural language, and translate that understanding into purposeful actions. Combined with World Models and advanced simulation, this approach enables machines to reason about their surroundings, anticipate changes, and make decisions with greater contextual intelligence.
This leap is more than technical, it is transformational. Robots equipped with these capabilities can adapt dynamically to complex, unpredictable environments, whether on a factory floor, in a warehouse, or out in the field. By fusing perception, language, and action into a single intelligent framework, Capgemini is setting the stage for autonomous systems that not only execute tasks, but also understand context, collaborate seamlessly, and unlock new possibilities for industries worldwide.
Edge technology powering Project REACH
This R&D runs on Intel’s advanced heterogeneous computing stack, delivering performance and flexibility at the edge:
- Local Inference on Intel® Core™ Ultra Processor Series 2:
Motion control executes on the CPU, while computer vision tasks can leverage the NPU or GPU, all on a single, compact chipset for maximum efficiency. - Geti + OpenVINO:
Accelerating AI development from training to deployment, this stack optimizes models for edge performance and rapid scalability. - Robotic Vision Control Framework:
Connecting perception to action, this layer integrates path planning, motion control, and ROS 2 for seamless robotic operations. - RealSense Cameras:
Providing depth perception, 3D mapping, and spatial awareness, these cameras enable robots to understand and navigate complex environments.
The real-world use cases driving adoption
Manufacturing
Robotics adoption in manufacturing is accelerating because AI-driven perception now makes it possible to automate tasks once considered too variable, such as orientation, inspection, and high-mix handling. With RealSense depth sensing and Geti-trained models, robots can detect defects, understand 3D orientation, and perform adaptive pick-and-place on Intel Core Ultra processors at the edge.
The impact is immediate: improved first-pass yield, reduced rework, and consistent 24/7 operations. Additionally, the modular Capgemini-Intel stack allows factories to retrain and redeploy models in hours, enabling scalable multi-line rollouts and lowering automation costs across complex production environments.
Logistics and Warehousing
In logistics, variability has long been the biggest barrier to automation: mixed parcels, unpredictable order flows, and constantly shifting inventory layouts make consistency a challenge. AI-enabled robotics powered by Intel’s edge computing stack help solve these challenges. Robots can now classify parcels, detect labels, measure dimensions, and route items dynamically with near-zero latency.
Beyond sorting, advanced perception enables palletizing and truck loading through intelligent, space-optimized stacking, reducing damage and improving utilization. Capgemini plans to integrate this research with warehouse management systems, creating autonomous workflows that boost throughput, minimize mis-sorts, and deliver 24/7 operational efficiency.
Healthcare Robotics
Healthcare robotics is entering a new era thanks in part to Intel’s deterministic edge computing, which delivers the precision and reliability required for surgical assistance, micro-manipulation, and rehabilitation, while maintaining local processing for reliability and privacy. AI-driven perception enables robots to identify surgical instruments, track movement in 3D, and assist clinicians with repetitive or fatigue-prone tasks, improving both safety and efficiency.
Capgemini’s engineering expertise ensures these systems meet regulatory-grade standards, making them safe, auditable, and ready for widespread clinical adoption. Together, these innovations help redefine what’s possible in patient care, bringing automation and intelligence to the heart of healthcare.
Agriculture and Field Inspection
Agriculture and infrastructure demand robotics that can operate autonomously across vast, unstructured, and often remote environments. Powered by RealSense™ depth perception and Intel edge computing, these robots can detect crop stress, inspect wind turbines, identify pipeline anomalies, and navigate rugged terrain, with minimal need for cloud connectivity.
The result? Lower costs and reduced risk compared to manual inspection, combined with improved accuracy and frequency of monitoring. This shift enables predictive, data-driven maintenance across farms, utilities, and industrial assets, transforming operations from reactive approaches to proactive, data-informed decision making and unlocking new efficiencies at scale.
Vision-Language-Action Robotics
Vision-Language-Action (VLA) robotics introduces a new paradigm where robots execute tasks based on natural language instructions combined with real-time visual understanding. This breakthrough makes automation accessible to non-technical workers, enabling intuitive human-robot collaboration. Capgemini integrates perception models, lightweight language models, and advanced motion planning into a unified system that understands intent and acts safely and intelligently.
Capgemini and Intel are not just showcasing a concept, they are delivering a fully engineered robotics stack that’s live, scalable, and evolving today. This demo offers a blueprint for industry-wide adoption, demonstrating how modular robotics, edge AI, and real-world autonomy can be deployed at scale to transform operations.
This is not robotics catching up with AI. This is robotics becoming AI-native.
