Don't Risk It!: Trajectory-Conditioned Distributional Risk for LiDAR-Only Crowd Navigation

Mathai Mathew Pulicken; Laksh Nanwani; Basant Sharma; Phani Teja Singamaneni; Arun Kumar Singh; K. Madhava Krishna

cs.ROOct 29, 2026

Don't Risk It!: Trajectory-Conditioned Distributional Risk for LiDAR-Only Crowd Navigation

Authors: Mathai Mathew Pulicken, Laksh Nanwani, Basant Sharma, Phani Teja Singamaneni, Arun Kumar Singh, K. Madhava Krishna

Organizations: Robotics Research Center, IIIT Hyderabad, India. · University of Tartu, Estonia. · Inria, Université de Lorraine, France.

Abstract

Efficient crowd navigation requires anticipating dynamic collision-free space without relying on fragile object tracking. We propose the Don't Risk It Planner (DRIP), an end-to-end LiDAR-only planning framework that couples occupancy forecasting with trajectory-conditioned clearance and uncertainty estimation. Operating directly on LiDAR-derived occupancy grid histories, DRIP predicts a distribution over minimum robot-obstacle clearance for candidate control sequences to evaluate collision risk. The model is trained using a risk cost that grounds predictive uncertainty directly in empirical collision outcomes rather than geometric regression alone. In closed-loop experiments with up to 45 pedestrians, DRIP achieves higher success rates with fewer collisions and reduced freezing behavior compared to state-of-the-art baselines.

Figures & tables

Fig. 2 : Overview of DRIP’s LiDAR-only prediction and trajectory-conditioned risk architecture. A history $\mathcal{H}$ of pose-warped BEV occupancy maps is encoded independently by a shared-weight spatial CNN, conditioned on ego velocity $\mathcal{V}$ through FiLM. Spatiotemporal Transformer blocks then reason over the temporal history and spatial BEV tokens to produce a shared scene representation $z_{scene}$ . Two branches use this representation: the occupancy decoder predicts the future evolution of the scene ( $\widehat{\mathbf{O}}_{k}$ ) over a $20$ -step horizon, while the trajectory-conditioned Distance Head combines scene features with candidate trajectory queries $\mathbf{q}_{k}$ to predict a stepwise clearance distribution $(\mu_{k},\sigma_{k}^{2})$ and a trajectory-specific kernel width $\lambda$ . These predictions are subsequently used to compute the distributional MMD 2 risk $\widehat{r}_{\boldsymbol{\theta}}(\mathbf{u})$ for each candidate trajectory $\mathbf{u}$ .

		Nominal Navigation (100 runs each)				Sensor Noise Degradation ( $\sigma=0.05\,\mathrm{m}$ ) (50 runs each)
Density	Planner	Succ. (%) $\uparrow$	Time (s) $\downarrow$	Path (m) $\downarrow$	Vel. (m/s) $\uparrow$	Succ. (%) $\uparrow$	Time (s) $\downarrow$	Path (m) $\downarrow$	Vel. (m/s) $\uparrow$
35 ped	DRLVO †	74.0	11.5	5.39	0.460	–	–	–	–
	DWA	79.0	13.5	5.08	0.355	38.0	31.8	7.33	0.195
	SCOPE	80.0	14.8	5.42	0.365	66.0	26.3	6.10	0.227
	TEB	71.0	12.3	5.46	0.413	30.0	40.2	18.09	0.368
	Ours	87.0	12.1	4.83	0.394	82.0	12.9	4.67	0.357

TABLE I : Closed-loop navigation performance and sensor noise robustness in the simulated Lobby world with 35 and 45 pedestrians. Higher success rate (

\uparrow

) and velocity (

\uparrow

) are better; lower completion time (

\downarrow

) and path length (

\downarrow

) are better. Best and second-best success rates are highlighted. ( † DRL-VO utilizes privileged pedestrian tracking/velocities and is excluded from direct sensor noise testing).

Fig. 3 : Box plot of best-of-

K

clearance residuals relative to privileged ground truth across dynamic evaluation episodes. The risk-calibrated model (( 18 )) yields a lower median error and narrower spread than the NLL-only baseline without

\mathcal{L}_{risk}

\mathcal{L}_{haz}

), demonstrating well-calibrated predictive variance that tightly bounds actual obstacle encounters.

Fig. 4 : Visualization of the occupancy forecasting. Top row: The first 5 steps of the predicted future occupancy maps ( $\widehat{\mathbf{O}}_{k}$ ), visualized as thresholded binary maps, taken from the full 20-step planning horizon. Bottom row: The corresponding ground-truth future occupancy maps ( $\mathbf{O}_{k}^{\mathrm{gt}}$ ). The network successfully anticipates the continuous motion of dynamic obstacles and maintains static boundary structures, closely matching the ground truth without relying on explicit pedestrian tracking.

Robot Navigation

Robot navigation studies how robots perceive, localize, plan, and move safely toward goals in static or changing environments.

See all 557 papers

Risk-Sensitive Crowd Navigation with Adaptive Ellipsoidal Conformal Prediction
Oct 5, 2026Ruihan A. Li, Ziyao Guo, Yingying LiDistribution Shift Robustness Robot Navigation
Coordinated Science Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL
Don't Freeze, Don't Crash: Extending the Safe Operating Range of Neural Navigation in Dense Crowds
Mar 6, 2026Jiefu Zhang, Yang Xu, Vaneet AggarwalSafe Robot Navigation Robot Navigation
Purdue University, USA
CoCoNav: Conformal Control for Safe Robot Navigation in Crowds
Aug 7, 2026Cheng Guo, Mingzhe Ni, Zheng Liang +5Constrained Motion Planning Conformal Prediction
Department of Computer Science, The University of Manchester, Manchester, UK. · Human-Robot Interfaces and Interaction Laboratory, Italian Institute of Technology, Genoa, Italy. · Genisom AI, Shanghai, China. +3

Oct 5, 2026J/K move · Enter open · S save

Risk-Sensitive Crowd Navigation with Adaptive Ellipsoidal Conformal Prediction

Ruihan A. Li, Ziyao Guo, Yingying Li

Coordinated Science Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL

Distribution Shift Robustness Robot Navigation Conformal Prediction Risk-Sensitive RL

Open paper AskPDF arXiv

Safe crowd navigation under distribution shift requires uncertainty representations that capture structured human-motion prediction errors and safety objectives that account for rare but consequential failures. Existing uncertainty-aware methods typically represent prediction errors using isotropic regions, which can be either overly conservative or poorly aligned with directional motion uncertainty. We introduce a risk-aware navigation framework that uses anisotropic conformal ellipsoids to translate structured prediction uncertainty into an episode-level conditional value-at-risk (CVaR) signal that regulates the Lagrangian safety penalty. In particular, adaptive ellipsoidal conformal prediction (AECP) captures directional prediction errors and adaptively calibrates uncertainty regions under distribution shift, while the resulting CVaR-regulated navigation policy is optimized using Lagrangian proximal policy optimization. We evaluate our proposed method under both in-distribution settings and out-of-distribution (OOD) settings involving shifts in pedestrian motion patterns. Compared with state-of-the-art baselines, our method maintains competitive in-distribution performance while improving success rates by 5.68-7.44 percentage points and reducing collision rates by 5.44-6.64 percentage points across OOD settings. We further deploy the trained policy without fine-tuning on a physical robot with onboard perception and CPU-only inference, showing that the full pipeline is feasible in physical crowd navigation.

Abstract

Robot Navigation

Explore similar work

Risk-Sensitive Crowd Navigation with Adaptive Ellipsoidal Conformal Prediction

Don't Freeze, Don't Crash: Extending the Safe Operating Range of Neural Navigation in Dense Crowds

CoCoNav: Conformal Control for Safe Robot Navigation in Crowds