GLST: Global-Local Spatio-Temporal Memory Decomposition Framework for Online Chained-Goal Navigation

Yousef Liang; Ryan Ardalan; Yarden Goraly; Chenhao Zhu; Yue Yin; Jonas Frey; Claire Tomlin; Marco Pavone

cs.ROOct 13, 2026

GLST: Global-Local Spatio-Temporal Memory Decomposition Framework for Online Chained-Goal Navigation

Authors: Yousef Liang, Ryan Ardalan, Yarden Goraly, Chenhao Zhu, Yue Yin, Jonas Frey, Claire Tomlin, Marco Pavone,

Organizations: Stanford University · University of California, Berkeley · NVIDIA

Abstract

Autonomous navigation in real-world environments often requires long-horizon operation under partial observability and evolving goals. Existing mapless navigation approaches predominantly assume fixed goals throughout an episode, limiting applicability to practical scenarios where goals change online during execution. We study online chained-goal navigation, where goals evolve through existing-goal redirection or new-goal emergence without explicit transition signals. Under this setting, we find navigation may require either continued exploration or strategic revisitation, creating conflicting memory requirements. We propose Global-Local Spatio-Temporal Belief Learning (GLST), a memory-decomposition framework that separates latent beliefs into global-local spatial and long-short temporal representations for long-horizon reasoning. We further introduce goal-conditioned rewards to balance exploration and exploitation under evolving goals. Extensive experiments in simulation across diverse terrains demonstrate improved chained-goal navigation performance over baselines. Real-robot experiments using the Spot quadruped further validate our memory decomposition framework.

Figures & tables

Figure 1: Overall framework and training pipeline of the proposed GLST method. (a) GLST framework showing the global-local spatial and temporal belief branches and actor-critic policy. Real-robot depth images or simulated height scans serve as shared inputs to both spatial branches. (b) Training pipeline including locomotion policy training, global spatial supervised pretraining, environment generation with chained-goal sampling, and navigation policy training.

Figure 2: Performance on chained-goal navigation. (a) Chained-goal success rate (SR), where empirical compounding is compared with independent-goal assumptions. (b) Single-goal SR and failure decomposition into timeout, collision, and excessive roll. (c) Navigation efficiency measured by SPL and optimal path length distributions. Increasing SPL across goals suggests improved efficiency under comparable task difficulty.

Figure 3: Visualization of learned spatial beliefs in two simulation environments: (1) maze and (2) pillars. (a) Global spatial beliefs reconstructed from predicted height, mask, and confidence maps, with robot trajectories overlaid. (b) Local observations obtained by fusing depth images from five cameras (front-left, front-right, left, right, and rear) into local height maps. (c) Simulation environments with the quadruped robot and executed trajectories overlaid.

Method	# Params (M)	Memory	Stats	Path Length 5 $\times$ [4,15] m					Path Length 5 $\times$ [4,20] m
				SR $\uparrow$	SPL $\uparrow$	TO $\downarrow$	Coll. $\downarrow$	Roll $\downarrow$	SR $\uparrow$	SPL $\uparrow$	TO $\downarrow$	Coll. $\downarrow$	Roll $\downarrow$
				(%)		(%)	(%)	(%)	(%)		(%)	(%)	(%)
Ours	$5.69$	GLST	Mean	$\underline{\mathbf{79.2}}$	$\underline{\mathbf{0.70}}$	$\underline{10.1}$	$\underline{9.1}$	$1.6$	$\underline{\mathbf{64.6}}$	$\underline{\mathbf{0.56}}$	$\underline{20.4}$	$\underline{12.9}$	$2.1$
Ours	$5.69$	GLST	Std	$0.8$	$0.01$	$0.6$	$0.9$	$0.5$	$1.8$	$0.02$	$1.1$	$0.7$	$0.3$
SRU [ 2 ]	11.8	[ 2 ]	Mean	$64.2$	$0.54$	$15.8$	$19.1$	$\underline{0.9}$	$50.3$	$0.42$	$21.2$	$27.9$	$\underline{0.6}$
SRU [ 2 ]	11.8	[ 2 ]	Std	$2.8$	$0.03$	$2.1$	$1.5$	$0.4$	$3.2$	$0.03$	$3.0$	$1.8$	$0.4$

Table 1: Navigation results over 3 independent runs of 1000-episode simulations. Failure modes include timeout (TO), collision (Coll.), and excessive roll (Roll). Best SR and SPL are bolded and underlined, while lowest failure rates are underlined.

Figure 4: Real-robot navigation experiments on the Spot quadruped across diverse indoor and outdoor environments. (1) Indoor experiments. (2) Outdoor forest experiments. (a) Experimental scenes with ghost robot overlays illustrating executed trajectories. (b) Inferred global spatial beliefs reconstructed from latent representations. (3) Long-horizon online chained-goal experiments with dead ends and dynamic obstacles. Additional chained-goal indoor and outdoor experimental results are provided in Appendix G .

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 5: Chained-goal success rate (%) as a function of goal index

n_{g}

, where

n_{g}=1,\dots,5

denotes the

n_{g}

-th goal in the sequence. The lower figures show the compounded success rate from 100% to the final success rate under the assumption of independent goal completion. The upper figures show the single-goal success rates reported in prior work, which are used to construct the chained-goal success rates. Data are adapted from (a) [ 2 ] , (b) [ 40 ] , (c) [ 41 ] , (d) [ 42 ] , and (e) [ 9 ] .

Figure 6: Detailed architecture of the proposed GLST framework. Observations are decomposed into global-local spatial and temporal belief representations. Global beliefs are recursively accumulated and refined through goal-conditioned attention, while fused belief representations are processed by a shared actor-critic policy. Auxiliary prediction heads supervise spatial belief learning through height, mask, and confidence reconstruction.

Path Length: $5\times[4,15]$ m
Terrain	Goal	Path length	# Episodes	Speed	SR (%) $\uparrow$	SPL $\uparrow$	Failure modes (%)
		(m)		(m/s)			Timeout $\downarrow$	Collision $\downarrow$	Roll $\downarrow$
Maze	Overall	$46.4\pm 0.7$	$1000$	$1.1$	$79.2\pm 0.9$	$0.70\pm 0.01$	$10.1\pm 0.7$	$9.1\pm 1.1$	$1.6\pm 0.5$
	G1	$9.3\pm 0.3$	$1000$	$1.0$	$90.6\pm 0.3$	$0.84\pm 0.01$	$5.0\pm 0.9$	$4.0\pm 0.8$	$0.4\pm 0.2$
	G2	$9.2\pm 0.3$	$906\pm 3$	$1.1$	$94.4\pm 0.5$	$0.85\pm 0.01$	$3.2\pm 0.2$	$2.0\pm 0.2$	$0.4\pm 0.2$
	G3	$9.4\pm 0.0$	$856\pm 5$	$1.1$	$96.5\pm 0.7$	$0.86\pm 0.01$	$1.9\pm 0.4$	$1.2\pm 0.4$	$0.2\pm 0.1$

Appendix

Table 2: Chained-goal navigation statistics across terrains and path-length settings. Results are aggregated over 3 runs of 1000-episode simulations and reported as mean

\pm

std across runs.

Method	# Params (M)	Memory	Stats	Path Length 5 $\times$ [4,15] m					Path Length 5 $\times$ [4,20] m
				SR $\uparrow$	SPL $\uparrow$	TO $\downarrow$	Coll. $\downarrow$	Roll $\downarrow$	SR $\uparrow$	SPL $\uparrow$	TO $\downarrow$	Coll. $\downarrow$	Roll $\downarrow$
				(%)		(%)	(%)	(%)	(%)		(%)	(%)	(%)
Ours	$5.7$	GLST	Mean	$\underline{\mathbf{79.8}}$	$\underline{\mathbf{0.77}}$	$\underline{7.9}$	$\underline{10.7}$	$1.6$	$\underline{\mathbf{64.2}}$	$\underline{\mathbf{0.61}}$	$\underline{12.1}$	$\underline{19.7}$	$4.0$
Ours	$5.7$	GLST	Std	$4.8$	$0.05$	$2.4$	$2.5$	$0.9$	$5.4$	$0.05$	$2.6$	$2.7$	$1.0$
SRU [ 2 ]	11.8	[ 2 ]	Mean	$64.7$	$0.54$	$15.0$	$18.9$	$\underline{1.4}$	$50.7$	$0.41$	$22.0$	$26.7$	$\underline{0.6}$
SRU [ 2 ]	11.8	[ 2 ]	Std	$2.9$	$0.03$	$2.2$	$1.4$	$0.4$	$3.3$	$0.03$	$2.7$	$1.9$	$0.3$

Appendix

Table 3: Navigation results over 1,000-rollout simulations (

\times 10

seeds). Failure modes include timeout (TO), collision (Coll.), and excessive roll (Roll). Best SR and SPL are bolded and underlined, while lowest failure rates are underlined.

Figure 7: Additional real-robot experiments: (a) single-goal navigation on stairs, (b) indoor navigation with three chained goals, and (c) forest navigation with three chained goals. Images are contrast-enhanced for visualization.

Memory-Augmented RL

See all 12 papers

GLAM: Training a latent world model over global spatiotemporal memory for active exploration and navigation
Sep 13, 2026I-Tak Ieong, Ruizhi Feng, Zhaoyang Lu +5Embodied Navigation Latent World Models
Tsinghua University · Lab for Brain-Inspired Embodied Intelligence, EBKernel Technologies Co., Ltd · Northeastern University +2
OVAL: Open-Vocabulary Augmented Memory Model for Lifelong Object Goal Navigation
Apr 14, 2026Jiahua Pei, Yi Liu, Guoping Pan +3Object-Goal Navigation Open-Vocabulary Object Navigation
Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China
STEGNav: Spatio-Temporal Event Graph Reasoning for Multimodal Lifelong Object Navigation
Aug 28, 2026Yang Chen, Zhenyu Huang, Wenbo Fu +4Vision-Language Navigation Object-Goal Navigation
State Key Laboratory of Novel Software Technology, Nanjing University · School of Intelligence Science and Technology, Nanjing University · School of Artificial Intelligence, Nanjing University

Sep 13, 2026J/K move · Enter open · S save

GLAM: Training a latent world model over global spatiotemporal memory for active exploration and navigation

I-Tak Ieong, Ruizhi Feng, Zhaoyang Lu, Yifei Cao+4

Tsinghua University · Lab for Brain-Inspired Embodied Intelligence, EBKernel Technologies Co., Ltd · Northeastern University+2

Embodied Navigation Latent World Models Embodied World Models World Model-Based Planning Object-Goal Navigation

Open paper AskPDF arXiv

Active exploration and semantic navigation require an embodied agent to build memory from partial observations, predict how the evolution of observed spatial memory may support future motion, and convert that prediction into actionable plans. We present GLAM, a goal-conditioned latent world model trained over global spatiotemporal memory, and GLAM NAV, the complete navigation system built around it. Given historical map tokens, a navigation goal, and the current robot pose, GLAM jointly predicts future map representations and robot-centric waypoint latents, allowing future spatial context and navigation intent to be inferred in a shared representation space. The model follows a JEPA-like latent prediction paradigm, operates directly on map-level latent tokens rather than RGB reconstruction, and uses a pretrained waypoint encoder-decoder to supervise and decode navigation plans within GLAM NAV. Training data are collected by replaying ObjectNav expert trajectories in Habitat over HM3D v0.2 scene assets and slicing them into multi-timescale prediction samples. On a controlled HM3D-ObjectNav subset reproduction setting, GLAM NAV improves over a reproduced BSC-Nav baseline in both success rate and success weighted by path length.

Abstract

Memory-Augmented RL

Explore similar work

GLAM: Training a latent world model over global spatiotemporal memory for active exploration and navigation

OVAL: Open-Vocabulary Augmented Memory Model for Lifelong Object Goal Navigation

STEGNav: Spatio-Temporal Event Graph Reasoning for Multimodal Lifelong Object Navigation