A model-based communication approach for distributed and connected vehicle safety systems

Information exchange for the purpose of creating situational awareness is the backbone of distributed systems that rely on communication. For example, connected vehicle safety systems rely on exchange of information between vehicles to allow each vehicle to create a real time map of its surrounding cars. These systems work by sampling the state of a physical process (vehicle trajectory) and communicating it to others. In this paper, we present a novel approach for communicating information of physical processes (such as vehicles) over a network using models of these processes. In this approach, traditional communication of data samples from a process is replaced by communication of models and model updates. This novel method allows each node in the distributed system to create a model-based view of its surrounding. The resulting model-based situational awareness map will have higher fidelity than only sample position information that is available through current methods. Using the example of a forward collision detection algorithm we demonstrate the significant performance improvement that is possible using this approach. We demonstrate that the system is able to maintain its performance at rates that are four times less than what current methods require.

I. Introduction

Distributed systems such as connected vehicle safety systems rely on communication amongst nodes (e.g., cars) as a means of creating situational awareness in the system [1]. Such situational awareness is the enabling factor in coordinating actions and sensing. Exchange of information in connected vehicles is usually done through vehicle-to-vehicle (V2V) communication [1] [3] [2]. In connected vehicle safety systems, similar to other distributed systems, each node (i.e., vehicles) frequently samples its own state and communicates the sample data to other nodes in its neighborhood. Each vehicle uses the received data from others to track them and to create a real time map of its surrounding. The tracking information is then used to determine collision hazards in real-time [3]. Given the real-time and safety nature of the applications, tracking must be done with relatively high precision, which results in high demands on the V2V network.