Authors : Sahasra V.; Yukaciny S.; Nithiyasri V.; Santhanalakshmi V.; Logeshwari; Dr. D. P. Sivasakti Balan; R. J. Thayumanaswamy

Volume/Issue : Volume 11 - 2026, Issue 10 - October

Google Scholar : https://tinyurl.com/5n7ws99e

DOI : https://doi.org/10.38124/ijisrt/26Oct455

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Abstract : Recent advances in computer vision, spatial computing, robotics, and embodied artificial intelligence (AI) have substantially improved machine capabilities in visual perception, navigation, spatial reasoning, and object manipulation. However, performance on structured benchmarks and controlled environments does not necessarily demonstrate equivalence to human spatial intelligence in complex real-world situations. This review examines whether current AI systems can reproduce the breadth and flexibility of human spatial and embodied intelligence, with particular emphasis on their implications for the potential replacement of humans by AI. The review synthesizes theoretical and empirical literature on human spatial cognition, embodied AI, vision-language-action models, robotics, and relevant AI capability benchmarks. The evidence is organised around four dimensions: (1) biological sensorimotor coupling and proprioception, (2) zero-shot spatial problem-solving and dynamic affordance recognition, (3) physical common-sense reasoning and adaptation, and (4) energy-efficient and context-sensitive spatial processing. Existing research indicates that AI systems have achieved considerable progress in perception, navigation, planning, and manipulation, but continue to encounter challenges involving unfamiliar environments, occlusion, uncertainty, dynamic obstacles, changing object configurations, physical interaction, long-horizon adaptation, and generalisation beyond training distributions. Human spatial intelligence, in contrast, integrates perception, proprioception, embodied experience, contextual knowledge, physical intuition, memory, and flexible real-time decision-making as an interconnected system. The reviewed evidence therefore suggests that current AI can augment and extend human spatial capabilities but cannot yet reproduce the breadth, adaptability, embodied experience, and contextual flexibility of human spatial intelligence. Spatial ability consequently represents an important dimension of human irreplaceability and supports a future paradigm of human–AI complementarity rather than complete human substitution.

Keywords : Artificial Intelligence, Human Intelligence, Spatial Intelligence, Embodied Intelligence, Robotics, Navigation, Manipulation, Spatial Reasoning, Generalisation, Human–AI Collaboration.

References :

1. OECD. (2025). Introducing the OECD AI Capability Indicators. OECD Publishing. The framework evaluates AI across nine human-capability domains, including vision, manipulation and robotic intelligence.
2. OECD. (2025). OECD AI Capability Indicators Technical Report. OECD Publishing. The technical report provides the methodology and evidence underlying the capability scales.
3. Fisher, R. B., Cohn, A. G., & Lochhead, C. (2025). Vision scale in Introducing the OECD AI Capability Indicators. OECD Publishing.
4. Messina, E. R. (2025). Manipulation scale in Introducing the OECD AI Capability Indicators. OECD Publishing.
5. Ho, C., Martin, R., Francis, J., & Oh, J. (2025). Robotic intelligence scale in Introducing the OECD AI Capability Indicators. OECD Publishing.
6. Mon-Williams, R. et al. (2025). Embodied large language models enable robots to complete complex tasks in unpredictable environments. Nature Machine Intelligence, 7, 592–601.
7. Phy-Q as a measure for physical reasoning intelligence. Nature Machine Intelligence. The study evaluates AI systems on physical reasoning and generalisation across physical scenarios.
8. PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-Agent Tasks. ICLR 2025. The benchmark evaluates spatial, temporal and collaborative constraints in human–robot tasks.
9. Raychaudhuri, S., & Chang, A. X. (2025). Semantic Mapping in Indoor Embodied AI—A Comprehensive Survey and Future Directions. The study reviews spatial mapping and reasoning challenges in unfamiliar indoor environments.
10. Brain-inspired spatial intelligence for embodied agents. Nature Communications (2026). The study examines spatial memory and generalisation as important components of embodied intelligence.
11. Jia, J. et al. (2026). Physics filtering favors the generalization of robot learning. npj Robotics. The study investigates robotic generalisation to unseen physical conditions.
12. Gui, S., Gui, K., & Luximon, Y. (2026). Explicit task reasoning empowering robotic manipulation. npj Artificial Intelligence. The study examines explicit reasoning for complex robotic manipulation.