CrossMaps: Confidence-Aware Open-Vocabulary Semantic Mapping for Rover Navigation

2026-06-15 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

CrossMaps is a real-time, confidence-aware open-vocabulary semantic mapping pipeline designed for rover navigation, constructing language-queryable maps from RGB-D data. This system enhances VLMaps-style approaches by integrating multi-scale CLIP embeddings with confidence-aware fusion and a dual-memory architecture comprising Short-Term Memory (STM) and Long-Term Memory (LTM). The STM actively aggregates noisy visual observations, leveraging geometric, semantic, and temporal confidence cues to refine data. Subsequently, confident and coherent cells are promoted to the LTM, establishing persistent semantic landmarks crucial for navigation. Engineered for deployment on a Jetson Orin-powered Unmanned Ground Vehicle (UGV) alongside SLAM, CrossMaps operates in real time, generating semantic heatmaps that can be queried using natural language to effectively guide rover movements.

Key takeaway

For Robotics Engineers designing autonomous navigation systems for rovers, CrossMaps offers a robust approach to real-time semantic mapping. You should consider integrating confidence-aware fusion and dual-memory architectures to improve map reliability and persistence. This enables language-queryable navigation, allowing your systems to interpret and act upon semantic commands more effectively. Evaluate its deployment on Jetson Orin-powered UGVs for enhanced operational autonomy.

Key insights

CrossMaps enables real-time, confidence-aware, open-vocabulary semantic mapping for rover navigation using a dual-memory architecture.

Principles

• Confidence-aware fusion improves map reliability.
• Dual-memory architecture separates transient from persistent data.
• Language-queryable maps enhance navigation.

Method

CrossMaps integrates multi-scale CLIP embeddings with confidence-aware fusion. It uses STM for noisy observations and LTM for persistent semantic landmarks, promoting confident cells from STM to LTM.

In practice

• Deploy on Jetson Orin-powered UGVs.
• Query maps with natural language.
• Guide rover navigation autonomously.

Topics

• Open-Vocabulary Semantic Mapping
• Rover Navigation
• CLIP Embeddings
• Dual-Memory Architecture
• Jetson Orin
• Real-time Perception

Best for: Computer Vision Engineer, Research Scientist, Robotics Engineer, AI Scientist, AI Engineer

Related on AIssential

• Heterogeneous LiDAR Early Fusion and Learned Re-Ranking Strategy for Robust Long-Term Place Recognition in Unstructured Environments — Heterogeneous LiDAR early fusion and learned re-ranking significantly boost place recognition in challenging unstructured agricultural environments.
• PSG-Nav: Probabilistic Scene Graph Navigation via Multiverse Decision Making — PSG-Nav uses probabilistic scene graphs and multiverse decision-making to navigate uncertain open-vocabulary environments.
• LongSpace: Exploring Long-Horizon Spatial Memory from Perception to Recall in Video — LongSpace enhances MLLMs' long-horizon spatial reasoning by integrating explicit 3D structural memory and question-guided retrieval.
• SemanticXR: Low Power and Real-time Queryable Semantic Mapping with an Object-Level Device-Cloud Architecture — SemanticXR uses an object-level device-cloud architecture for low-power, real-time, queryable semantic mapping in XR.
• ByteDance Introduces Astra: A Dual-Model Architecture for Autonomous Robot Navigation — ByteDance's Astra uses a dual-model architecture for robust, general-purpose robot navigation in complex indoor spaces.
• LiteVLA-H: Dual-Rate Vision-Language-Action Inference for Onboard Aerial Guidance and Semantic Perception — Edge VLA latency is pre-fill dominant, necessitating dual-rate scheduling for reactive guidance and semantic awareness.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.