Multi-cam Multi-map Visual Inertial Localization: System, Validation and Dataset (2412.04287v1)

Abstract: Map-based localization is crucial for the autonomous movement of robots as it provides real-time positional feedback. However, existing VINS and SLAM systems cannot be directly integrated into the robot's control loop. Although VINS offers high-frequency position estimates, it suffers from drift in long-term operation. And the drift-free trajectory output by SLAM is post-processed with loop correction, which is non-causal. In practical control, it is impossible to update the current pose with future information. Furthermore, existing SLAM evaluation systems measure accuracy after aligning the entire trajectory, which overlooks the transformation error between the odometry start frame and the ground truth frame. To address these issues, we propose a multi-cam multi-map visual inertial localization system, which provides real-time, causal and drift-free position feedback to the robot control loop. Additionally, we analyze the error composition of map-based localization systems and propose a set of evaluation metric suitable for measuring causal localization performance. To validate our system, we design a multi-camera IMU hardware setup and collect a long-term challenging campus dataset. Experimental results demonstrate the higher real-time localization accuracy of the proposed system. To foster community development, both the system and the dataset have been made open source https://github.com/zoeylove/Multi-cam-Multi-map-VILO/tree/main.

• The paper presents a novel VILO system that integrates multi-camera arrays and multiple maps for real-time, drift-free localization in autonomous robots.
• It introduces new evaluation metrics that address the limitations of traditional VINS and SLAM approaches by enabling causal and accurate pose estimation.
• The extensive nine-month dataset, collected in a dynamic campus environment, robustly validates the system's performance under challenging conditions.

Multi-cam Multi-map Visual Inertial Localization: System, Validation and Dataset

The paper presents a novel approach to visual-inertial localization specifically aimed at addressing the challenges encountered by autonomous robots during navigation, particularly concerning real-time, causal, and drift-free localization. The proposed system incorporates a multi-camera, multi-map methodology, along with a suite of tools for evaluation and validation, supported by an extensive dataset collected over a long-term period.

In the field of autonomous robotics, precise map-based localization is instrumental for reliable path planning and execution. Traditional VINS (Visual-Inertial Navigation Systems) and SLAM (Simultaneous Localization and Mapping) methodologies, while effective, suffer from inherent drawbacks such as drift accumulation in VINS and post-processed non-causal trajectory adjustments in SLAM. These limitations impede their integration within the control loops of robotic systems which necessitate real-time positional feedback.

The presented research introduces an advanced multi-cam multi-map visual inertial localization (VILO) system. This system is designed to deliver accurate, real-time, and drift-free position feedback by leveraging a multi-camera array and a strategy for handling multiple maps without necessitating global consistency among them. The research identifies three core aspects to tackle:

1. System Design: The authors emphasize the flexibility and modularity of the system which allows seamless integration of multiple isolated maps. Unlike conventional map-reliant systems that mandate prior map fusion, the VILO system permits real-time switching and integration of maps based on sector-specific needs.
2. Evaluation Metrics: Recognizing the inadequacy of existing metrics such as ATE and RPE in evaluating drift-free localization, the paper proposes new metrics to robustly assess localization performance within a map frame accurately.
3. Dataset Collection and Evaluation: To substantiate the system's claims, the authors provide an extensive dataset gathered by deploying a state-of-the-art multi-camera and inertial measurement unit (IMU) hardware configuration. This dataset, characterized by challenging migrations such as appearance and structural changes, spans a campus environment over nine months, providing a robust platform for testing and validation of VILO systems.

Crucially, this research highlights the importance of causal localization wherein current pose estimates do not rely on future information, aligning with real-world applications where timely and accurate feedback is essential for robot control.

The implications of this research are twofold:

• Practical Applications: The introduction of a robust VILO system translates into enhanced autonomous capabilities for robots, especially in unstructured environments. The flexibility with multi-map capabilities means that systems can be deployed in diverse settings without extensive pre-mapping, ensuring operational readiness and adaptability.
• Theoretical Contributions: From a computational standpoint, the paper introduces deterministic global optimal solutions for pose estimation, enhancing reliability in environments with high outlier rates. This methodological contribution can significantly influence future research in spatial localization and mapping technologies.

Looking forward, the open-sourcing of both the system and its dataset stands to significantly further community-driven research and innovation in robotic localization. Additionally, the principles elucidated in this research could inform the development of more adaptive, robust, and scalable navigation systems across various autonomous platforms. As AI and robotics continue to integrate into complex, dynamic environments, such advancements in localization will undoubtedly play a pivotal role in their efficacy and adoption.