The University of Edinburgh SENSOPAC Site

Simulation Environment
Relates to SENSOPAC Work Package 2: 2C.2
People involved: Djordje Mitrovic, Stefan Klanke, Sethu Vijayakumar

Role within SENSOPAC

We develop and run a physics-based simulation environment that allows us to work with a dynamically and kinematically consistent virtual copy of the SENSOPAC robot, e.g. for evaluating control strategies and learning algorithms.

When developing new methods and algorithms for controlling robots and evaluating sensor information, or more generally for dealing with any piece of hardware, it is highly beneficial to have an accurate simulation of the hardware at one's disposal for various reasons:

Safety and cost effectiveness: If faulty control strategies lead to a crash, any damage is only virtual.
One can evaluate algorithms before the real hardware is built, that is, hardware and software engineers can work in parallel. This is especially important for collaboration within the SENSOPAC project, e.g. for plugging in our control approaches as soon as the hardware of the integrated hand-arm system (DLR) is ready.
Shortcomings of the system may be detected and be incorporated in design corrections.
Real hardware signals are always noisy. Within a simulation, the level of noise can be controlled, so the robustness of e.g. control algorithms can be assessed in detail.
Some algorithms may need heavy optimization before being applicable in a real-time environment. In the simulation, efficiency can be taken into account at a later stage.

Our simulation environment is based upon the Open Dynamics Engine (ODE) for physics simulation and OpenGL for visualization. We can simulate articulated rigid bodies that are composed of simple geometric primitives, but also complicated objects given their trimesh model.

Above: A screenshot of our simulation environment with the DLR lightweight arm.

Links

Website of the Open Dynamics Engine (ODE)

Subproject overview