← Back ICRA 2026

NeuroLiDAR: Adaptive Frame Rate Depth Sensing Via Neuromorphic Event-LiDAR Fusion

Darshana Rathnayake, Dulanga Kaveesha Weerakoon Weerakoon Mudiyanselage, Meera Radhakrishnan, Archan Misra

PDF

AI summary

Key figure (auto-extracted from paper)

Fusing low-frame-rate LiDAR with neuromorphic event cameras enables adaptive, high-effective frame rate depth sensing (~66 Hz) with ~29% lower error, without the power costs of continuous high-speed scanning.

Adaptive frame rate Event-LiDAR fusion Neuromorphic vision Depth extrapolation Embedded robotics ELiDAR dataset

Problem

Commercial LiDARs are limited to low frame rates (5–10 Hz) due to power and hardware constraints, leaving critical gaps in tracking rapid scene changes. Conversely, fixed high frame rates waste energy, creating a need for a responsive, adaptive depth sensing paradigm.

Approach

NeuroLiDAR dynamically fuses sparse LiDAR depth maps with continuous, low-power event camera streams. A lightweight event-driven detector triggers depth extrapolation only during significant scene changes, synthetically boosting the effective frame rate on demand.

Key results

Achieves effective depth sensing rates up to ≈66 Hz on embedded hardware
Reduces depth reconstruction RMSE by ≈29% versus conventional 10 Hz LiDAR
Introduces ELiDAR, a synchronized simulated and real-world event-LiDAR benchmark dataset
Demonstrates real-time, low-latency deployment on NVIDIA Jetson Orin devices

Why it matters

Provides a scalable, energy-efficient pathway for autonomous vehicles and robots to achieve high-temporal-fidelity perception in dynamic environments without costly hardware upgrades.

Abstract

LiDARs are widely used for 3D depth recon- struction, but their performance is often limited by inherent hardware constraints that impose trade-offs between range, spatial resolution, and frame rate. Many LiDAR systems typically operate at low frame rates (e.g., 5-10 Hz), prioritizing long-range sensing over responsiveness to rapid scene changes. We present NeuroLiDAR, an adaptive depth sensing framework that achieves effective frame rates of up to ≈66 Hz by fusing temporally sparse LiDAR data with temporally dense inputs from neuromorphic event cameras. NeuroLiDAR integrates two components: event-based keyframe detection and event-guided depth extrapolation, to dynamically adjust the sensing rate in response to scene dynamics. To evaluate our approach, we intro- duce ELiDAR, a dataset spanning outdoor and indoor scenarios, and show that NeuroLiDAR reduces depth reconstruction error by ≈29% in RMSE while achieving adaptive frame rates between 27.8–47.3 Hz. Our code and dataset are available at https://github.com/darshanakgr/neurolidar.

Index terms

Sensor Fusion Deep Learning for Visual Perception RGB-D Perception