CAVISE Overview

CAVISE (Connected & Automated Vehicle Integrated Simulation Environment) combines driving simulation, traffic simulation, communication and perception experiments in one workspace.

Component	Responsibility
CARLA	3D world, vehicle physics and sensors.
OpenCDA	Scenario orchestration, driving logic, agent services and evaluation.
SUMO	Traffic simulation synchronized with CARLA when co-simulation is enabled.
Artery	Network simulation connected to OpenCDA through CAPI.
OpenCOOD	Cooperative-perception models and multi-agent inference.
Models	Checkpoint and AdvCP runtime asset bundles fetched as needed.

Start with CARLA and OpenCDA, then add capabilities as the experiment requires. Installation prepares the environment. First Simulation walks through a bounded first run.

How the Components Connect

OpenCDA connects to CARLA’s API and advances the scenario. With --cosim, it also connects to a running SUMO server over TraCI. With --with-capi, it exchanges communication data with Artery. The base Artery CAPI scenario connects to that same SUMO server. CAPI’s Protobuf modules are generated by the build, so these experiments require a Protobuf-capable OpenCDA image.

Cooperative perception uses OpenCOOD and a model checkpoint. AdvCP hooks into that perception pipeline. Attacks against behavior services use a separate attack framework. The OpenCDA Overview page describes OpenCDA capabilities and shows scenario demonstrations.

History

OpenCDA and Artery were developed independently. In 2023 and 2024, the CAVISE team implemented CAPI integration around the original realistic_town06_cosim experiment. Current launch instructions use the shared artery/scenarios/capi configuration. See SUMO and Artery.