SkyAgent: Revisable Target Memory for Continuous Aerial Search

Kaibo Yang; Xingyu Wu; Enze Zhu; Yajie Chai; Xuyang Li; Jiahui Cui; Xiaoxuan Liu; Lei Wang

cs.ROOct 16, 2026

SkyAgent: Revisable Target Memory for Continuous Aerial Search

Authors: Kaibo Yang, Xingyu Wu, Enze Zhu, Yajie Chai, Xuyang Li, Jiahui Cui, Xiaoxuan Liu, Lei Wang,

Organizations: Aerospace Information Research Institute, Chinese Academy of Sciences · Key Laboratory of Target Cognition and Application Technology (TCAT) · School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences · Baidu Inc.

Abstract

Continuous aerial search builds target hypotheses from sparse, viewpoint-dependent observations. When later evidence contradicts a stored hypothesis, however, the candidate can remain available to the planner and continue triggering approach or verification actions. We call this trajectory-level failure false candidate persistence (FCP). We equip SkyAgent with Revisable Target Memory (RTM), which associates evidence with persistent candidates, discounts correlated views, and revises each candidate through Tentative, Confirmed, and Rejected states. The key interface binds every generated waypoint to its source candidate: rejection therefore invalidates stale goals, while conservative reactivation protects temporarily occluded targets. Across four splits from the CityAVOS, UAV-ON, and OpenUAV benchmarks, our SkyAgent equipped with RTM outperforms all competitors, improving SR by 2.90--6.31 points while substantially reducing Behavior-FCP and post-contradiction travel with a low false-rejection rate. Across 35 matched real-world DJI scenarios per method, RTM consistently improves task success and behavioral reliability, further validating its state-to-action correction under physical deployment conditions.

Figures & tables

Figure 1: False candidate persistence in continuous aerial search. A static distractor looks like the target from an early viewpoint ( A ) and is contradicted by an independent later view ( B ). An append-only state keeps $C_{1}$ active; the drone wastes route on a known-wrong candidate ( C , top). RTM rejects $C_{1}$ , cancels its pending waypoint, and resumes search toward the true target without erasing the audit trail ( C , bottom).

Figure 2: Revisable Target Memory. Candidate Evidence Memory groups sequential observations by independent viewpoint and accumulates signed evidence for each persistent candidate. Revise-and-Suppress updates each candidate’s lifecycle through a state-dependent transition and propagates rejection to the planner by invalidating every pending waypoint sourced from the rejected candidate.

Benchmark	Method	SR $\uparrow$	Eff. $\uparrow$	Dist. $\downarrow$	BFCPR $\downarrow$	Waste $\downarrow$	FRR $\downarrow$
CityAVOS-Hard (SkyAgent $n{=}380$ )	RE ( Ji et al. 2026 )	5.05	2.47	674.16	N/A	N/A	N/A
	FBE ( Ji et al. 2026 )	8.07	6.34	628.70	N/A	N/A	N/A
	STMR ( Ji et al. 2026 )	20.96	13.23	547.77	N/A	N/A	N/A
	PRPSearcher (Hard subset) ( Ji et al. 2026 )	29.62	16.65	507.60	N/A	N/A	N/A
	Append-only SkyAgent	33.95	23.16	431.13	45.18	120.76	N/A
	SkyAgent + RTM (Hard)	40.26	30.14	349.14	15.27	44.21	2.91

Table 1: Performance on aerial benchmarks. SkyAgent rows are paired internal comparisons. External SR and Eff. values are reported by the cited papers; Dist. values are outputs of our benchmark reproduction rather than values copied from their tables. External methods may use different perception front ends. SR, Eff., BFCPR, and FRR are percentages; Dist. and Waste are meters.

Figure 3: Mechanism audit layout. Panel A contrasts append-only and RTM trajectories after the reference contradiction at $t^{\star}$ . Panel B aligns signed evidence, candidate state, and stale-goal cancellation. Panel C defines the distance-quartile summary used for the supplementary stress analysis.

Variant	SR $\uparrow$	SFCPR $\downarrow$	BFCPR $\downarrow$	Waste $\downarrow$	FRR $\downarrow$	ORR $\uparrow$	Latency $\downarrow$
Append-only	33.95	55.33	45.18	120.76	–	77.08	–
Verify current action	36.05	50.25	38.19	96.17	1.96	80.21	6.53
RTM w/o negative evidence	35.53	51.24	39.80	103.18	0.99	82.29	7.29
RTM w/o viewpoint grouping	37.63	27.23	21.29	60.92	8.65	77.08	2.37
Revision state only	38.42	19.90	28.86	74.24	3.88	89.58	8.72
Permanent rejection	38.68	15.42	12.44	38.26	11.65	65.63	2.52

Table 2: Mechanism ablations on CityAVOS-Hard (

n{=}380

). Closed-loop rates use row-specific externally annotated events; the shared-stream analysis and all event counts are reported separately. Waste is meters and Latency is steps.

Figure 4: Representative real-flight revision sequence. Inspection rejects $C_{1}$ and cancels its queued waypoint; search resumes toward $C_{2}$ , confirmed from an independent view. Table 3 reports all paired trials.

Method	SR $\uparrow$	Dist. $\downarrow$	BFCPR $\downarrow$	FRR $\downarrow$	Rev.Lat.(s) $\downarrow$	Int. $\downarrow$
SPF (TF)	54.29	275.68	50.00	12.50	6.31	17.14
PSC-AVDN (TF)	60.00	257.94	40.91	16.67	5.24	11.43
Append-only SkyAgent	57.14	264.37	45.45	–	–	14.29
SkyAgent + RTM	71.43	223.81	13.64	4.17	2.78	5.71

Table 3: Matched DJI flights. Results use 35 matched scenarios/method (140 trials). Rates are percentages, Dist. is meters, and Rev.Lat. is seconds; TF denotes no task-specific training or fine-tuning.

Aerial Robotics

Aerial robotics studies the design, perception, control, and coordination of flying robots for navigation, inspection, transport, and physical interaction.

See all 326 papers

SLIDER: Sparse History-Guided Aerial Robot Target Search using Sliding Local Maps
Jul 12, 2026Xiaolei Hou, Zheng Pan, Hua Lan +6Aerial Robotics Robot Navigation
Towards Lifelong Aerial Autonomy: Geometric Memory Management for Continual Visual Place Recognition in Dynamic Environments
Apr 10, 2026Xingyu Shao, Zhiqiang Yan, Liangzheng Sun +5Cross-View Geo-Localization Visual Place Recognition
Online Planning for Sparse Ground Target Search from a High-Altitude UAV under Partial Observability
Oct 1, 2026Ashik E Rasul, Hyung-Jin YoonActive Perception Remote Sensing Small Object Detection
Department of Mechanical and Nuclear Engineering Tennessee Technological University Cookeville, USA

Jul 12, 2026J/K move · Enter open · S save

SLIDER: Sparse History-Guided Aerial Robot Target Search using Sliding Local Maps

Xiaolei Hou, Zheng Pan, Hua Lan, Zhenghao Zou+5

Aerial Robotics Robot Navigation Frontier-Based Exploration Topological Mapping

Open paper AskPDF arXiv

Efficient exploration and target search in large-scale unknown environments remain challenging for aerial robots due to the demands of broad spatial coverage, fine-grained perception, and real-time decision-making. This paper presents SLIDER, a lightweight and memory-efficient framework that avoids reliance on globally dense maps by combining a local sliding map with sparse global history information. A novel observation quality evaluation method is proposed, leveraging historical poses and sensor models to assess point cloud data in real-time, enabling efficient frontier detection. To support scalable and responsive planning, an incremental viewpoint clustering strategy dynamically adapts to local updates, significantly reducing the number of candidate targets and decreasing computational load. A sparse global topological map is incrementally maintained to assist global planning and cost evaluation. Extensive simulations and real-world experiments demonstrate that the proposed system outperforms state-of-the-art methods in memory usage, decision latency, and search efficiency.

Abstract

Aerial Robotics

Explore similar work

SLIDER: Sparse History-Guided Aerial Robot Target Search using Sliding Local Maps

Towards Lifelong Aerial Autonomy: Geometric Memory Management for Continual Visual Place Recognition in Dynamic Environments

Online Planning for Sparse Ground Target Search from a High-Altitude UAV under Partial Observability