/Reliable Navigation for Autonomous Vehicles in Connected Vehicle Environments by using Multi-agent Sensor Fusion

Reliable Navigation for Autonomous Vehicles in Connected Vehicle Environments by using Multi-agent Sensor Fusion

Suryansh Saxena
Master's Thesis, Tech. Report, CMU-RI-TR-17-45, Robotics Institute, Carnegie Mellon University, August, 2017

Download Publication (PDF)

Copyright notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author’s copyright. These works may not be reposted without the explicit permission of the copyright holder.


The age of intelligent vehicles is here, and with the rise in autonomous navigation, we face a new challenge in finding ways to increase the safety and reliability of navigation for autonomous vehicles in dynamically changing environments.

To date, most methodologies in autonomous navigation mainly rely on the data feed from local sensors. However, erroneous measurements in sensor data might reduce the overall safety and confidence in navigation. These errors may be persistent and/or temporal, and give rise to false predictions made by the autonomous agents, usually classified as either a false positive or a false negative. Though false positives may influence navigation safety in an indirect manner, false negatives directly and significantly affect it.

In this thesis, we explore a methodology for sharing and fusing sensor data of close-proximity autonomous or intelligent vehicles to identify and minimize false negatives in local environment interpretation. For each agent, the methodology compares the local SLAM maps with maps generated by other close proximity agents to identify false negatives in local interpretation.

As per the methodology, each autonomous agent simultaneously localizes and maps its local environment. This map, in turn, is encoded into a low-resolution message and shared over the DSRC communication protocol. Next, the agents distributively fuse this information together to build world interpretation. Each agent then statistically analyses its own interpretation with respect to the world interpretation for the common regions of interest. The proposed statistical algorithm outputs the measure of similarity between local and world interpretations and identifies false negatives (if any) for the local agent. This measure, in turn, can be used to inform the agents to change their kinematic behavior in order to account for the errors in local interpretation. Finally, each agent records the measure and instances of erroneous interpretations, which over a period of time helps analysis and quantifies the sensor health.

We evaluate the qualitative and quantitative efficacy of our proposed methodology and algorithms using a widely accepted traffic simulator called SUMO in scenarios that have a high accident frequency. We use ROS in tandem with SUMO to simulate and quantify the behavior of autonomous vehicles. Finally, we also propose a communication architecture to enforce the proposed methodology in the real world and quantify the efficacy of the communication architecture by conducting latency testing.

BibTeX Reference
author = {Suryansh Saxena},
title = {Reliable Navigation for Autonomous Vehicles in Connected Vehicle Environments by using Multi-agent Sensor Fusion},
year = {2017},
month = {August},
school = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-17-45},
keywords = {Autonomous Vehicles, Connected Vehicles, DSRC, Multiagent Coordination},