The Four Strategic Pillars of UM-PolyU CREI
UM-PolyU CREI has four major research directions (strategic pillars), ensuring comprehensive coverage of the robotic value chain.
Serving as the consciousness nexus for robotics, Robotics Intelligence integrates perception, cognition, decision-making, planning, and control to instill machines with a generative understanding of physical laws, affordances, and causality. Its value lies in enabling robots to reason about their actions and adapt to novel situations, forming the scientific cornerstone for generalizable autonomy. Current pioneering approaches, such as world models that construct internal representations to understand and predict future states, vision-language-action models that ground instructions in physical tasks, reinforcement learning that optimises behaviour through environmental interaction, and spatial intelligence that enables reasoning about geometry and manipulation, provide concrete pathways toward this embodiment. By addressing the critical limitations of current systems, this pillar ensures robots can learn, infer, and plan within the unstructured complexities of the real world.
Acting as the foundational carrier for robotics, this pillar integrates embodiment, advanced sensors, specialized chips, and high-performance computing to provide a stable and reliable physical base for widespread deployment of robotics intelligence in the real world. It establishes a shared, standardized platform that enables machines to physically interact with and adapt to dynamic environments, serving as the practical cornerstone for scalable and generalizable autonomy. Current research advances this infrastructure through diverse frontiers, including cloud computing for scalable data processing and collaborative intelligence, edge computing for real-time decision-making, micro-robotics that access and interact within confined or high-risk environments, and soft robotics for compliant and adaptive manipulation. By providing the essential “body” for the intelligent “mind,” Perception and Embodied Computing constitutes the vital pathway for transforming virtual intelligence into physical agents.
Functioning as the execution guarantor, this pillar ensures rigorous verification, validation, safety, and reliability across the entire robotic lifecycle. Recognizing that scalable autonomy depends on systematic assurance rather than ad-hoc checks, we elevate safety from a peripheral constraint to an intrinsic, system-level attribute. Its strategic value lies in establishing an engineering assurance framework for AI-enabled autonomy, covering performance-guaranteed AI functions, robustness analysis of AI enabled systems, and AI-in-the-loop system analysis and design, grounded in systems engineering for AI-empowered robotic platforms. This framework ensures that local reliability aggregates into global system assurance through continuous state assessment, fault diagnosis, and compliance verification, supported by transparent metrics and auditable evidence trails, with clear pathways for certification. Ultimately, it provides the foundation of trust required for robots to operate in highly sensitive and critical scenarios.
Serving as the adaptability amplifier for robotics, this pillar expands the scope of practical application by enhancing both operational flexibility and system intelligence. Its value lies in forging a robust bridge between fundamental research and real-world implementation, leveraging adaptive intelligence to reconfigure complex workflows, integrate with digital twins, and respond to dynamic environmental demands. This work is applied across several key areas. In the public sector, robots enhance service quality by guiding passengers, supporting elderly care, and performing inspections. In logistics, the use of predictive analytics allows robots to coordinate warehouse operations, optimize routing, and ease supply chain bottlenecks. Furthermore, the focus on precise operation enables automated diagnostics, surgical assistance, and patient monitoring, while AI-driven sensing supports high-resolution terrain mapping, conduct environmental monitoring, strengthen disaster preparedness, and improve infrastructure design. As a transformation engine that generates immediate social and scientific impact, this pillar simultaneously refines fundamental research questions, ensuring the Centre’s long-term relevance and establishing a sustainable pathway from scientific discovery to profound societal benefit.
