Abstract

In recent years, many new vehicles equipped with autonomous driving capabilities have been deployed around the world, but they still have a number of failure cases. One proposed solution to improve performance in these scenarios is cooperative perception, a paradigm in which vehicles communicate with other vehicles and nearby sensors to collect additional perception information about the environment. Prior work has demonstrated that cooperative perception can be accomplished with the high accuracy and low latency needed for autonomous driving. However, state-of-the-art systems rely on unrealistic assumptions about the conditions of the network connections between participating nodes and the compute resources available on individual nodes. In this paper, we propose a new approach to cooperative perception that adapts to the state of network and compute resources. We demonstrate its advantages over static approaches with traces representing a variety of scenarios. By making cooperative perception robust to changes in the network and compute resources, we push it a step further toward viable real-world deployment.

Publication Date

4-14-2026

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Science (MS)

Department, Program, or Center

Computer Science, Department of

College

Golisano College of Computing and Information Sciences

Advisor

Fawad Ahmad

Advisor/Committee Member

Zachary Butler

Advisor/Committee Member

Richard Lange

Campus

RIT – Main Campus

Download

Share

COinS