Concept

CommonRoad-Control closes the gap between motion planning and control research for autonomous driving. Many planning benchmarks and open-source projects do not consider the intertwined nature of planning and control. Only when analyzed with a closed-loop state-of-the-art controller and a dynamics simulation can motion planning algorithms be truely evaluated.

Another major gap we aim to close with that project, is the lack of open-source control toolboxes that are fully compatible with planning benchmarks yet modular enough to easily integrate novel controllers or motion planners.

Finally, we provide a dedicated numerical simulation using ODE-solvers and kinematic or dynamic models of different fidelity to simulate the application of the controller outputs to the plant, including noise and disturbances.

Easy use with the Easy API

The Easy API of our toolbox allows for the seemless plug-and-play of our implemented controllers. It is targeted at motion planning researchers that already use CommonRoad and its planners and simply want to test their planning results with an out-of-the-box controller and simulation.

Modularity

The control loop and the interdependence of its constituting parts make it notoriously difficult to implement in a modular fashion. Our toolbox offers modules for all major parts of the control loop, such as:

• model-free and model-based controllers;
• the dynamics simulation;
• kinematic and dynamic vehicle dynamics models;
• uncertainties (sensor noise and disturbances);
• sensor models (e.g., full state feedback); and
• the planner integration.

Through the use of interfaces (= base classes), we give users the ability to integrate their own modules fairly easily.

Integration of your own work

We offer detailes tutorials on how the integrate your own motion planner, controller or vehicle model in these tutorials. You can also follow our long examples for the PID controller and the MPC and replace our modules with your own work.