Self in Space: Benchmarking Self-Awareness and Spatial Cognition in UAV Embodied Intelligence

2026-07-14

AI Summary

Current intelligent drone systems generally lack explicit modeling of their own states, overemphasizing environmental understanding while neglecting self-awareness, thereby struggling to support embodied intelligence in complex real-world scenarios. To address this limitation, this work proposes a unified “self–space” cognitive framework and introduces SIS-Bench, the first evaluation benchmark integrating self-awareness with spatial understanding. SIS-Bench comprises 13 tasks and 4,856 structured question-answer pairs, assessing multimodal large models across perception, memory, and reasoning through a task-conditioned construction pipeline and expert validation. We further enhance dynamic self-modeling by incorporating an optical flow–visual feature fusion representation for motion perception. Experiments demonstrate that our approach significantly improves models’ perceptual and mnemonic capabilities in spatial cognition and self-awareness, and effectively generalizes to downstream decision-making tasks.

Abstract

Autonomous UAV systems increasingly rely on multimodal large language models (MLLMs) to operate in complex real-world environments. Such embodied scenarios require not only understanding the surrounding space but also maintaining a coherent representation of the agent itself. However, existing UAV-oriented approaches and benchmarks remain largely environment-centric, primarily focusing on spatial understanding tasks, with the agent's self-awareness remaining implicit. To address this gap, we introduce SIS-Bench, a benchmark for evaluating embodied spatial intelligence in UAV scenarios under a unified self-in-space formulation. SIS-Bench organizes evaluation along two complementary dimensions, space and self, and a three-level hierarchy of perception, memory, and reasoning. It contains 4,856 question--answer pairs across 13 tasks derived from 1,646 real-world UAV videos through a task-conditioned construction pipeline with expert verification.Extensive evaluations reveal that current MLLMs exhibit fundamental limitations in modeling dynamic and agent-centered processes. In particular, we observe a clear imbalance between spatial cognition and self-awareness, as well as a progressive performance degradation across cognitive levels.Motivated by these findings, we further explore a motion-aware representation that incorporates self-related dynamics through optical flow and visual feature fusion. Experimental results show that modeling agent motion consistently improves perception and memory performance, not only in spatial cognition but also in self-awareness, and generalizes to downstream UAV decision-making tasks.Our results highlight the importance of self-awareness for advancing embodied spatial intelligence, and provide both a new benchmark and empirical evidence for motion-aware self-in-space modeling.

Problem

• self-awareness
• spatial cognition
• embodied intelligence
• UAV
• benchmark

Innovation

• self-awareness
• spatial cognition
• embodied intelligence
• motion-aware representation
• UAV benchmark