[GB/*] Context-Adaptive Robot Motion Planning and Control

When programming industrial robots using the Robot Operating System, there is no systematic way to ensure context-dependent safety behaviour. An example is to restrict the movement parameters of the robot, i.e., the maximum velocity and acceleration, during runtime. This is a prerequisite, when switching between human-aware/human-friendly "slow" modes and efficient independent "fast" modes. One problem is, that this behaviour must be ensured in several places, including the robot drivers (there are multiple for different kinds of movements) and the motion planning subsystem (which includes kinematic solvers). These components must be analysed systematically and extended accordingly. Using the resulting adaptive motion planner and robot driver, a simple adaptive use case should be built.
Requirements:

• Some knowledge of C++ (the code is C++, although fairly simply written)
• Basic knowledge of C/C++ development with cmake and Linux
• A motivation to understand and extend complex software systems.

Task:

• Investigate the structure of the ROS planning and execution pipeline and the dataflow of the robot models within it.
• Investigate how model parameters are used in the drivers and motion planning.
• Extend both of these components to support configurable during runtime.
• Implement an adaptive demo scenario (this is mostly provided, it just has to be extended to use the new adaptation mechanism).

Betreuer: Johannes Mey; Sebastian Ebert