Work

TACO: Temporal Consensus Optimization for Continual Neural Mapping

Year: 2026

Type: conference-paper

Abstract: without storing or replaying previous data.Through extensive simulated and real-world experiments, we show that TACO robustly adapts to scene changes, and consistently outperforms other continual learning baselines.The code can be found at https://iconlab.negarmehr.com/TACO

Source: Robotics: Science and Systems

DOI: 10.15607/rss.2026.xxii.045

Authors: Xunlan Zhou, Hongrui Zhao, Negar Mehr

Institutions: Nanjing University, University of Illinois Urbana-Champaign, University of California, Berkeley

Language: English

Topic: Robotics and Sensor-Based Localization

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: bronze

Funders: National Science Foundation, Nvidia, Division of Electrical, Communications and Cyber Systems

Awards: CPS-Small: Provable Enforcement of Hard Constraints in Reinforcement Learning-Based Controllers for Safety-Critical CPS, CAREER: Socially-Aware Control of Autonomy: Reshaping Urban Mobility in Traffic Networks with Mixed Vehicle Autonomy