Knowledge Representation and Reasoning for Robotics and Autonomous Agents in AI-Empowered Intelligent Engineering Systems: A Review

Abstract

Knowledge representation and reasoning (KR&R) constitute the cognitive foundation enabling robotics and autonomous agents to operate reliably within complex, dynamic environments characteristic of intelligent engineering systems. Anchored in systems engineering theory, cyber-physical systems (CPS) principles of bidirectional physical-cyber coupling, and information-theoretic measures of uncertainty and information gain, this review synthesizes advances that transform raw sensor streams into structured, actionable knowledge. Ontologies, description logics, and neuro-symbolic architectures are examined alongside learning-augmented reasoning mechanisms that support adaptive planning and explainable decision-making. Concrete exemplars illustrate translation into practice: a digital twin of an automotive assembly cell employing reinforcement learning for real-time task rescheduling under production disturbances, and a semantic knowledge-graph framework guiding a collaborative manipulator through context-aware grasp selection and failure recovery. Benefits in operational resilience, human-robot synergy, and lifecycle optimization are weighed against limitations in computational tractability, brittleness under distributional shift, and certification challenges for safety-critical reasoning. An original insight proposes augmenting CPS reference architectures with semantically layered entropy-aware belief maintenance to optimize knowledge fidelity versus update cost. The review concludes by outlining research trajectories toward hybrid KR&R frameworks that reconcile symbolic rigor with data-driven adaptability, thereby advancing trustworthy autonomy in next-generation intelligent engineering systems.