Robust Execution of Abstract Task Plans on Mobile Robots
- 0241 - 80 21 003
Solving a planning problem is itself not sufficient to use a planning system on an actual robotic system. Instead, additional platform requirements need to be taken into account, the execution of the plan needs to be monitored, and abstract plans must be transformed into an action sequence that is executable on the particular robot. One of the goals of my thesis is to provide declarative platform models of robotic system components. Based on these models, we can formulate platform constraints that need to be satisfied during execution. The proposed framework transforms a given abstract task plan into an executable action sequence that takes the additional constraints into account. Additionally, the model can be used during execution to detect and recover from any component failures and unexpected changes in the environment. As foundation for this execution framework, we will provide a temporal extension of the situation calculus that allows formulating quantitative temporal constraints. The task plan and the additional constraints are then translated into a constraint satisfaction problem, whose solution is used to generate the executable action sequence. The framework will be able to execute an abstract task plan, that is, a plan that was generated without a specific robotic system in mind, on a particular robotic system, given the system‘s platform model and constraints.