DUGM-R: Uncertainty-Aware Dynamic Grid Mapping and Risk-Triggered Recovery for Learned Local Navigation

Haoyun Feng; Adrian Rubio-Solis; Zhaodong Guo; George Mylonas

cs.ROSep 23, 2026

DUGM-R: Uncertainty-Aware Dynamic Grid Mapping and Risk-Triggered Recovery for Learned Local Navigation

Authors: Haoyun Feng, Adrian Rubio-Solis, Zhaodong Guo, George Mylonas

Organizations: The Hamlyn Centre for Robotic Surgery, Imperial College London, London, United Kingdom. · Department of Computing, Imperial College London, London, United Kingdom.

Abstract

Learned local navigation in crowded indoor environments is sensitive to how dynamic obstacle motion is represented, while collision-prone behaviour may persist after nominal policy training. We present a risk-aware reinforcement-learning framework that addresses these two issues through an uncertainty-aware Dynamic Uncertainty Grid Map (DUGM) and a modular post-training recovery mechanism. DUGM combines local occupancy, estimated obstacle motion, and motion-estimation uncertainty in a robot-centric representation. After the nominal policy is frozen, a finite-horizon Risk Value Function (RVF) is trained from nominal rollouts and used to trigger a dedicated recovery policy when continued nominal execution is predicted to be collision-prone. Experiments in a held-out NVIDIA Isaac Sim clinical-logistics benchmark show that uncertainty-aware dynamic representation improves nominal navigation over static and deterministic alternatives, while the recovery mechanism further mitigates residual collision-prone behaviour. The complete framework is also deployed directly on a TurtleBot3 without policy fine-tuning, retraining, or site-specific adaptation, retaining the performance trend observed in simulation. These results indicate that uncertainty-aware dynamic representation and post-training recovery provide complementary mechanisms for improving learned local navigation.

Figures & tables

Fig. 1: Overview of the proposed framework: (a) DUGM observation construction, (b) RVF-triggered runtime switching, and (c) the two-stage training pipeline.

Fig. 2: Architecture template for the nominal policy, recovery policy, and RVF. Each network has an independent instance of blocks 1–4 and its corresponding output heads; no parameters are shared across networks.

Panel of Figure 3 — Fig. 3: RVF scores queried across robot positions in a fixed scene. Green/red indicate low/high scores; blue arrows show motion directions. Top-left insets specify the robot velocity used in each panel.

Panel of Figure 4 — Fig. 4: Held-out Isaac Sim clinical-logistics benchmark used for final evaluation, including static furnishings and dynamic pedestrian interactions.

Group	Method	SR $\uparrow$	CR $\downarrow$	TR $\downarrow$	$\bar{v}$ (m/s) $\uparrow$
External	DWA	51.1%	15.6%	33.3%	0.150
	Nav2 MPPI	35.6%	3.3%	61.1%	0.114
	ORCA	27.8%	65.6%	6.6%	0.125
	DRL-VO	38.9%	61.1%	0.0%	0.134
	OGM–Lagrangian	58.9%	41.1%	0.0%	0.140
Observation	OGM	57.8%	41.1%	1.1%	0.149

TABLE I: Simulation results over the 90 held-out benchmark episodes.

Fig. 5: Observation ablation: colored curves show robot trajectories for the indicated representations; white curves show pedestrian paths. Matching inset borders identify observations at locally comparable interaction states, sampled at different times.

Fig. 6: DUGM trajectories without recovery (magenta) and with RVF-triggered recovery (colored by RVF score: green, low; red, high). White curves show pedestrian paths.

Panel of Figure 8 — Fig. 7: Representative views of the eight real-robot evaluation scenes with overlaid robot–pedestrian interactions at selected timestamps. Green annotations denote the robot, while yellow annotations denote pedestrians.

Method	SR $\uparrow$	CR $\downarrow$	TR $\downarrow$
DWA	62.5%	8.3%	29.2%
OGM	70.8%	29.2%	0.0%
DUGM	79.2%	20.8%	0.0%
DUGM + recovery	83.3%	4.2%	12.5%

TABLE II: Real-robot evaluation results over 24 runs per method.

Occupancy Grid Mapping

See all 17 papers

DiffCVaR: Reinforcement Learning for Risk Adaptation via Differentiable CVaR Barrier Functions
May 20, 2026Xinyi Wang, Taekyung Kim, Bardh Hoxha +2Control Barrier Functions Risk-Sensitive RL
Department of Robotics, University of Michigan, Ann Arbor, MI, 48109, USA · Toyota Motor North America, Research & Development, Ann Arbor, MI, 48105, USA · Department of Aerospace Engineering, University of Michigan, Ann Arbor, MI, 48109, USA
DODGER: Safety-Guided Reinforcement Learning for Robot Navigation Among Dynamic Obstacles
Sep 30, 2026Sanghyuk Park, Kwanwoo Lee, Taekyung Kim +2RL for Robotics Control Barrier Functions
Department of Intelligence and Information, Seoul National University, Republic of Korea · Department of Robotics, University of Michigan, Ann Arbor, MI, USA · Department of Mechanical Engineering, Yonsei University, Republic of Korea +1
UCON: Uncertainty-aware Navigation with Historical Re-association in Dynamic Environments
Sep 24, 2026Bing Sun, Yue Lin, Yongsheng Yuan +3Trajectory Optimization Robot Navigation
All authors are with Dalian University of Technology, Dalian 116024.China.

May 20, 2026J/K move · Enter open · S save

DiffCVaR: Reinforcement Learning for Risk Adaptation via Differentiable CVaR Barrier Functions

Xinyi Wang, Taekyung Kim, Bardh Hoxha, Georgios Fainekos+1

Department of Robotics, University of Michigan, Ann Arbor, MI, 48109, USA · Toyota Motor North America, Research & Development, Ann Arbor, MI, 48105, USA · Department of Aerospace Engineering, University of Michigan, Ann Arbor, MI, 48109, USA

Control Barrier Functions Risk-Sensitive RL Safe RL Conditional Value-at-Risk Robot Collision Avoidance

Open paper AskPDF arXiv

Planning through crowded environments under uncertain obstacle motions remains difficult, as stochastic interactions often induce overly conservative behavior or reduced efficiency. To address this challenge, we propose an end-to-end risk adaptation framework for crowd navigation under obstacle-motion uncertainty modeled by a Gaussian mixture model. The framework combines reinforcement learning~(RL) with a differentiable quadratic-program safety layer based on Conditional Value-at-Risk~(CVaR) barrier functions, jointly learning nominal control input, risk level, and safety margin and enforcing explicit probabilistic safety constraints. This design enables context-aware adaptation, promoting efficient behavior while invoking caution only when necessary. We conduct extensive evaluations in dynamic, uncertain, and crowded environments across varying obstacle densities and robot models, and further assess generalization under three out-of-distribution cases. Comparisons across optimization-based, RL-based, and integrated RL and optimization methods are provided, and the proposed method is shown to deliver the strongest overall performance in safety, efficiency, and generalization under uncertainty.

Abstract

Occupancy Grid Mapping

Explore similar work

DiffCVaR: Reinforcement Learning for Risk Adaptation via Differentiable CVaR Barrier Functions

DODGER: Safety-Guided Reinforcement Learning for Robot Navigation Among Dynamic Obstacles

UCON: Uncertainty-aware Navigation with Historical Re-association in Dynamic Environments