Message Passing Neural Network Versus Message Passing Algorithm for Cooperative Positioning

Abstract

Cooperative Positioning (CP) relies on a network of connected agents equipped with sensing and communication technologies to improve the positioning performance of standalone solutions. In this paper, we develop a completely data-driven model combining Long Short-Term Memory (LSTM) and Message Passing Neural Network (MPNN) for CP, where agents estimate their state from inter-agent and state-dependent measurements. The proposed LSTM-MPNN model is derived from a parallelism with the probability-based Message Passing Algorithm (MPA) for CP, from which the graph-based structure of the problem and message passing scheme is inherited. In our solution, the LSTM block predicts the motion of the agents, while the MPNN elaborates the node and edge embeddings for an effective inference of the agent's state. We present numerical evidence that our approach can enhance position estimation, while being at the same time an order of magnitude less complex than typical particle-based implementations of MPA for non-linear problems. In particular, the presented LSTM-MPNN model can reduce the error on agents' positioning to one third compared to MPA-based CP, it holds a higher convergence speed and better exploits cooperation among agents.

Deep Learning-Based Cooperative LiDAR Sensing for Improved Vehicle Positioning

“…Recent works exploit the capabilities of NN over networks, also known as Message Passing Neural Network (MPNN) [51], [52], which learn the correct association directly from semantic data [38]. MPNNs can discern both linear and non-linear relations between input and output data, they are time scalable similar to SPA [53] and have demonstrated superior performance on loopy graphs, provided the availability of sufficient training data [54], [55].…”