Skip to content

About ADORe

ADORe (Automated Driving Open Research) is an open-source software framework for testing and developing automated driving – on real vehicles, in simulation, and in combination with intelligent road infrastructure. It is developed mainly at the German Aerospace Center (DLR), but is explicitly intended for everyone to use, extend, and embed in their own projects.

At its core, ADORe is not a “finished” self-driving product. Instead, it is a research and experimentation platform:

  • A set of modular libraries for automated driving (planning, control, multi-agent coordination, interfaces, etc.).
  • A system-level framework that connects vehicles, simulations, and infrastructure.
  • An environment to try out new ideas, compare strategies, and run experiments under realistic conditions.

What problems ADORe tries to solve

Modern automated driving isn’t just about making one clever car. It’s about how many vehicles and the surrounding infrastructure work together as a system:

  • Vehicles need to drive safely and comfortably on their own.
  • They should also be able to cooperate with other vehicles and roadside units.
  • Researchers need a way to evaluate algorithms at this system level, not just as isolated components.

ADORe is built to support exactly this kind of work:

  • Single-vehicle autonomy – local planning and control on the vehicle itself.
  • Multi-agent / infrastructure-supported automation – coordination via V2X (vehicle-to-everything) communication, infrastructure assistance, and backend services.
  • End-to-end experiments – from high-level mission goals down to trajectory tracking and feedback.

Key ideas

1. Modular building blocks, not a monolith

ADORe is organised as a set of ROS 2–based components that can be combined and swapped:

  • Core libraries for geometry, motion planning and trajectory tracking.
  • Nodes for single-vehicle automation (SAAD).
  • Nodes and services for multi-agent / infrastructure-based automation (MAAD).
  • Interfaces towards sensors, vehicles, simulators, and backends.

You can:

  • Use individual libraries inside your own stack.
  • Replace ADORe components with your own, as long as you respect the interfaces.
  • Embed ADORe modules into existing projects without adopting everything.

The idea is that you can start small (e.g. just use a planning or control component) and grow into more of the framework over time.

2. Designed for both vehicles and infrastructure

ADORe is built around the assumption that automation will often be shared between vehicle and infrastructure, rather than living only inside the car:

  • On-board stack for local autonomy on research vehicles.
  • Infrastructure-side components that monitor traffic, compute cooperative plans, or assist vehicles via V2X.
  • Back-end middleware to connect ADORe to project-specific cloud or edge services.

This makes it possible to run experiments such as:

  • Infrastructure-guided trajectories in complex intersections.
  • Cooperative manoeuvres between several vehicles.
  • Remote-supported operation and supervision.

3. Works in simulation and on real vehicles

ADORe is meant to be portable across environments:

  • Integration with traffic and vehicle simulators (e.g. SUMO, CARLA and related tools) for large-scale or repeatable tests.
  • Deployment on DLR’s research vehicles with real sensors and actuators.
  • Containerised setups so that the same software can run in CI, on dev machines, and on cars with minimal change.

This allows workflows like:

  1. Prototype an idea purely in simulation.
  2. Move the same ADORe components to a vehicle, keeping interfaces and behaviour consistent.
  3. Use infrastructure-side ADORe instances to coordinate both simulated and real traffic participants in mixed scenarios.

Who ADORe is for

Although ADORe is developed at DLR, it is meant for a wider community:

  • Research groups working on planning, control, cooperative driving, V2X, traffic management, remote operation, or safety concepts.
  • Industry teams who need a flexible, inspectable framework for experiments or prototyping.
  • Students and hobbyists who want to explore automated driving beyond single-vehicle demos.

The software is open source, and the goal is to make it possible to:

  • Reproduce and extend research results.
  • Plug in your own algorithms (planners, controllers, decision-makers).
  • Integrate ADORe into larger toolchains or platforms.

What ADORe is not

To set expectations clearly:

  • ADORe is not a ready-to-deploy commercial driverless system.
  • It does not target full SAE Level 4+ product deployments, but rather research, prototyping, and demonstrations.
  • It assumes expert users who are comfortable working with ROS 2, Docker, and modern C++/Python.

Safety, validation, and certification for production systems remain the responsibility of whoever uses the framework.

Typical ways people use ADORe

Here are a few common patterns:

  • Use it as a full stack on a research vehicle or in simulation, and tune or swap selected modules.
  • Embed individual libraries (e.g. motion planning, trajectory tracking) inside your own ROS 2 system.
  • Connect your infrastructure (traffic controllers, roadside units, digital twins) to ADORe’s multi-agent layer and run cooperative experiments.
  • Prototype new ideas (e.g. new decision logic, cooperative strategies, remote support concepts), while reusing all the “boring but necessary” parts like map handling, routing, and interfaces.

Learn more

If you want to dive deeper into the architecture and research background behind ADORe, see the VEHITS 2025 paper “ADORe: Unified Modular Framework for Vehicle and Infrastructure-Based System Level Automation” included in this repository.

For the practical side (how to build, run, and extend the framework), check the developer-oriented documentation and package-level READMEs in the codebase.