On the Redundancy Detection in Keyframe-Based SLAM

Abstract

Egomotion and scene estimation is a key component in automating robot navigation, as well as in virtual reality applications for mobile phones or head-mounted displays. It is well known, however, that with long exploratory trajectories and multi-session mapping for long-term autonomy or collaborative applications, the maintenance of the everincreasing size of these maps quickly becomes a bottleneck. With the explosion of data resulting in increasing runtime of the optimization algorithms ensuring the accuracy of the Simultaneous Localization And Mapping (SLAM) estimates, the large quantity of collected experiences is imposing hard limits on the scalability of such techniques. Considering the keyframe-based paradigm of SLAM techniques, this paper investigates the redundancy inherent in SLAM maps, by quantifying the information of different experiences of the scene as encoded in keyframes. Here we propose and evaluate different information-theoretic and heuristic metrics to remove dispensable scene measurements with minimal impact on the accuracy of the SLAM estimates. Evaluating the proposed metrics in two state-of-the-art centralized collaborative SLAM systems, we provide our key insights into how to identify redundancy in keyframe-based SLAM.

Distributed Formation Estimation Via Pairwise Distance Measurements

“…While these approaches are demonstrated to achieve accurate co-localization, they are designed for a manageable number of agents, and will not achieve real-time performance when applied to several tens or hundreds of agents. Some recent works [12], [13] have investigated how SLAM graphs can be reduced in order to boost the scalability of collaborative SLAM, demonstrating significant reductions in the data necessary to represent an environment, yet still not allowing collaborative SLAM algorithms to scale up to swarm size. Another popular approach to simplify relative localization among UAVs is to attach a salient, pre-known and unique pattern (termed marker) on each agent.…”

“…These error terms enforce the constraints in Eq. (12). Special care has to be taken for the consensus term involving the rotation.…”

Information-Driven Direct RGB-D Odometry

“…Mutual information between two random variables. It measures how much knowing one of the variables reduces the uncertainty about the other [21]:…”

“…Redundancy detection using Mutual Information. As in [21], the redundancy ψ(K j ) of a keyframe with respect to the others can be expressed by…”

COVINS: Visual-Inertial SLAM for Centralized Collaboration

“…At the same time, multi-agent SLAM poses significant challenges, such as accurate co-localization, ensuring consistency with multiple agents simultaneously contributing information, and managing scalability with regard to the large amount of contributed data. Several works have shown good progress tackling one ore more of these challenges over the last few years [3,11,12,18,19]. However, collaborative SLAM is a relatively young research field, albeit a very promising one.…”