Graph based Cooperative Localization for Connected and Semi-Autonomous Vehicles

Abstract

Cooperative Real-time Localization is expected to play a crucial role in various applications in the field of Connected and Semi-Autonomous vehicles (CAVs), such as collision avoidance/warning, cooperative adaptive cruise control, etc. Future 5G wireless systems are expected to enable costeffective Vehicle-to-Everything (V2X) systems, allowing CAVs to share the measured data with other entities of the network. Typical measurement models usually deployed for this problem, are absolute position from Global Positioning System (GPS), relative distance to neighboring vehicles and relative angle or azimuth angle, extracted from Light Detection and Ranging (LIDAR) or Radio Detection and Ranging (RADAR) sensors. In this paper, we provide a cooperative localization approach that performs multi modal-fusion between the interconnected vehicles, by representing a fleet of connected cars as an undirected graph, encoding each vehicle position relative to its neighboring vehicles. This method is based on the so called Laplacian Processing, a Graph Signal Processing tool, that allows to capture intrinsic geometry of the undirected graph of vehicles rather than their absolute position on global coordinate system, significantly outperforming current state of the art approaches, in terms of localization mean square and maximum absolute error and computational complexity.

Graph Laplacian Extended Kalman Filter for Connected and Automated Vehicles Localization

“…While the aforementioned CL schemes offer significant benefits, they are mainly evaluated using trajectories generated by simplified kinematic models and more importantly they focus only on the pair-wise measurements, without considering the topology of a large number of CAVs that form undirected graphs with varying topologies. Although in our previous study in non-Bayesian Centralized and Distributed Laplacian Localization (CLL and DLL) [9], we proved the significance of exploiting this type of connectivity using the graph Laplacian operator, the tracking properties were neglected. Therefore, in this work we focus on proposing novel cooperative tracking schemes which apart from measurements, take also into account the underlying graph of involved vehicles, by efficiently utilizing the graph Laplacian operator.…”

Graph Laplacian Diffusion Localization of Connected and Automated Vehicles

“…Finally, we constructed the Cumulative Distribution Function (CDF) of LMSE. As it will be shown, all three proposed approaches outperform in terms of location estimation accuracy, the distributed low cost variant Distributed Laplacian Localization (DLL) [11] of CLL and the method of [18], named Maximum Likelihood based Localization (MLL). The former estimates only the ego-vehicle location, using the noisy positions of local neighborhood and the socalled local graph Laplacian operator.…”

“…each vehicle relying only to its neighbors to be in place to estimate the entire common location vector of the VANET which belongs. Our previous work on non-Bayesian Centralized Laplacian Localization (CLL) [11], [12], [13], which linearly correlates via the graph Laplacian operator the multi-modal measurement modalities along with the connectivity representation of vehicles, facilitated the design of the proposed approaches. Therefore, in this work we focus on proposing novel diffusion localization schemes for CAVs, which aim to converge to CLL utilizing the graph Laplacian operator.…”

Empowering cyberphysical systems of systems with intelligence

“…The main aspect of [31], as well as [30], is that vehicles must reach a consensus about features state, in order to improve their location. Graph Laplacian CL has been introduced in [32], [33]. Centralized or distributed Laplacian Localization formulates a LS optimization problem, which fuses the heterogeneous inter-vehicular measurements along with the V2V connectivity topology through the linear Laplacian operator.…”