Scalable Factor Graph-Based Heterogeneous Bayesian DDF for Dynamic Systems

Heterogeneous Bayesian decentralized data fusion captures the set of problems in which two robots must combine two probability density functions over non-equal, but overlapping sets of random variables. In the context of multi-robot dynamic systems, this enables robots to take a "divide and conquer" approach to reason and share data over complementary tasks instead of over the full joint state space. For example, in a target tracking application, this allows robots to track different subsets of targets and share data on only common targets. This paper presents a framework by which robots can each use a local factor graph to represent relevant partitions of a complex global joint probability distribution, thus allowing them to avoid reasoning over the entirety of a more complex model and saving communication as well as computation costs. From a theoretical point of view, this paper makes contributions by casting the heterogeneous decentralized fusion problem in terms of a factor graph, analyzing the challenges that arise due to dynamic filtering, and then developing a new conservative filtering algorithm that ensures statistical correctness. From a practical point of view, we show how this framework can be used to represent different multi-robot applications and then test it with simulations and hardware experiments to validate and demonstrate its statistical conservativeness, applicability, and robustness to real-world challenges.

Towards Decentralized Heterogeneous Multi-Robot SLAM and Target Tracking

In many robotics problems, there is a significant gain in collaborative information sharing between multiple robots, for exploration, search and rescue, tracking multiple targets, or mapping large environments. One of the key implicit assumptions when solving cooperative multi-robot problems is that all robots use the same (homogeneous) underlying algorithm. However, in practice, we want to allow collaboration between robots possessing different capabilities and that therefore must rely on heterogeneous algorithms. We present a system architecture and the supporting theory, to enable collaboration in a decentralized network of robots, where each robot relies on different estimation algorithms. To develop our approach, we focus on multi-robot simultaneous localization and mapping (SLAM) with multi-target tracking. Our theoretical framework builds on our idea of exploiting the conditional independence structure inherent to many robotics applications to separate between each robot's local inference (estimation) tasks and fuse only relevant parts of their non-equal, but overlapping probability density function (pdfs). We present a new decentralized graph-based approach to the multi-robot SLAM and tracking problem. We leverage factor graphs to split between different parts of the problem for efficient data sharing between robots in the network while enabling robots to use different local sparse landmark/dense/metric-semantic SLAM algorithms.

Heterogeneous Bayesian Decentralized Data Fusion: An Empirical Study

In multi-robot applications, inference over large state spaces can often be divided into smaller overlapping sub-problems that can then be collaboratively solved in parallel over `separate' subsets of states. To this end, the factor graph decentralized data fusion (FG-DDF) framework was developed to analyze and exploit conditional independence in heterogeneous Bayesian decentralized fusion problems, in which robots update and fuse pdfs over different locally overlapping random states. This allows robots to efficiently use smaller probabilistic models and sparse message passing to accurately and scalably fuse relevant local parts of a larger global joint state pdf, while accounting for data dependencies between robots. Whereas prior work required limiting assumptions about network connectivity and model linearity, this paper relaxes these to empirically explore the applicability and robustness of FG-DDF in more general settings. We develop a new heterogeneous fusion rule which generalizes the homogeneous covariance intersection algorithm, and test it in multi-robot tracking and localization scenarios with non-linear motion/observation models under communication dropout. Simulation and linear hardware experiments show that, in practice, the FG-DDF continues to provide consistent filtered estimates under these more practical operating conditions, while reducing computation and communication costs by more than 95%, thus enabling the design of scalable real-world multi-robot systems.