Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

DOOR-SLAM: Distributed, Online, and Outlier Resilient SLAM for Robotic Teams (1909.12198v2)

Published 26 Sep 2019 in cs.RO

Abstract: To achieve collaborative tasks, robots in a team need to have a shared understanding of the environment and their location within it. Distributed Simultaneous Localization and Mapping (SLAM) offers a practical solution to localize the robots without relying on an external positioning system (e.g. GPS) and with minimal information exchange. Unfortunately, current distributed SLAM systems are vulnerable to perception outliers and therefore tend to use very conservative parameters for inter-robot place recognition. However, being too conservative comes at the cost of rejecting many valid loop closure candidates, which results in less accurate trajectory estimates. This paper introduces DOOR-SLAM, a fully distributed SLAM system with an outlier rejection mechanism that can work with less conservative parameters. DOOR-SLAM is based on peer-to-peer communication and does not require full connectivity among the robots. DOOR-SLAM includes two key modules: a pose graph optimizer combined with a distributed pairwise consistent measurement set maximization algorithm to reject spurious inter-robot loop closures; and a distributed SLAM front-end that detects inter-robot loop closures without exchanging raw sensor data. The system has been evaluated in simulations, benchmarking datasets, and field experiments, including tests in GPS-denied subterranean environments. DOOR-SLAM produces more inter-robot loop closures, successfully rejects outliers, and results in accurate trajectory estimates, while requiring low communication bandwidth. Full source code is available at https://github.com/MISTLab/DOOR-SLAM.git.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (5)
  1. Pierre-Yves Lajoie (12 papers)
  2. Benjamin Ramtoula (9 papers)
  3. Yun Chang (43 papers)
  4. Luca Carlone (109 papers)
  5. Giovanni Beltrame (64 papers)
Citations (169)

Summary

An Overview of DOOR-SLAM: Distributed, Online, and Outlier Resilient SLAM for Robotic Teams

The paper introduces DOOR-SLAM, a notable advancement in the domain of multi-robot systems, focusing on a Simultaneous Localization and Mapping (SLAM) approach that emphasizes distributed operation, online processing, and robust outlier rejection. Unlike conventional SLAM systems that often rely on centralized processing or necessitate complete connectivity among robots, DOOR-SLAM represents a concerted effort to overcome these limitations and presents a scalable alternative tailored for collaborative tasks executed by robotic teams.

Key Aspects and Contributions

The central objective of DOOR-SLAM is to facilitate a shared understanding of the environment among robots while efficiently managing localization without depending on external systems like GPS. This capability is particularly crucial in GPS-denied environments such as subterranean exploration. DOOR-SLAM is robust against perception outliers, an Achilles' heel in many existing systems, and employs less conservative parameters without compromising on accuracy.

Main Contributions:

  1. Distributed Architecture: The system operates through peer-to-peer communication and gracefully adjusts even if robots are not fully connected. This architectural choice mitigates bandwidth stress and enhances efficiency in communication-restricted environments.
  2. Outlier Rejection Mechanism: The system integrates a distributed pose graph optimizer alongside a pairwise consistent measurement set maximization algorithm. This innovation aids in effectively dismissing false loop closure candidates, improving the reliability of trajectory estimations.
  3. Data-efficient Front-end: To minimize raw data exchange, DOOR-SLAM generates inter-robot loop closures leveraging NetVLAD descriptors. This component significantly reduces communication overhead while maintaining performance.

Experimental Evaluations

DOOR-SLAM was thoroughly evaluated through simulations, benchmark datasets like KITTI, and real-world field trials, including GPS-denied subterranean environments. The results indicate that DOOR-SLAM consistently produces accurate trajectory estimates while successfully filtering out outliers, which can severely distort mapping tasks.

Notable Results:

  • In simulated environments, DOOR-SLAM's robust outlier rejection mechanism demonstrated a marked improvement in the consistency of loop closures.
  • Field tests showed a significant reduction in data transmission when the system was compared against centralized SLAM methods, emphasizing its communication efficiency.
  • Implementations on drones confirmed the system's capability under practical constraints, validating its real-time application potential.

Theoretical and Practical Implications

From a theoretical standpoint, DOOR-SLAM contributes to the body of research on distributed SLAM by demonstrating that precision and robustness can be achieved without centralized data processing, thus paving the way for more adaptive and scalable SLAM solutions. Practically, its robust outlier rejection enables teams of robots to accurately map and navigate challenging environments collaboratively. This development offers significant potential advancements in industries ranging from autonomous navigation in logistics to exploration in hazardous or inaccessible terrains.

Future Prospects

The research opens up interesting avenues for future work. There is room to further enhance the robustness of the DOOR-SLAM system by exploring resistance to packet drops and correlated groups of outliers, thus ensuring enhanced safety and resilience under varying conditions. Additionally, extending DOOR-SLAM to accommodate a broader range of sensors and robotic configurations can expand its applicability across diverse real-world scenarios.

In summary, DOOR-SLAM represents a sophisticated advancement in multi-robot SLAM systems, offering a data-efficient, robust, and flexible architecture that meets the unique demands of real-world robotic applications. It underscores the practicability and theoretical expansion possibilities within the field of distributed SLAM, potentially redefining how robots perceive and interact with their surroundings in complex environments.

Github Logo Streamline Icon: https://streamlinehq.com
Youtube Logo Streamline Icon: https://streamlinehq.com