SimPy Model for fast testing

[ZuyuanZhu]

This is an extension of @marc-hanheide 's idea(#56) to model topological nodes as simpy.Container that can each hold a robot, ensuring robots cannot occupy the same spot and therefore allow to detect deadlocks.

Features of this package:

• Independent of ROS
• Quick test on a Topo map for multi-robot path planning and multi-picker task allocating (next target)

Assumptions:

• Robot requests a new node once he reaches the current node
• Robot release the previous node once he reaches halfway towards the next node

TODO:

• [X] Robot avoids occupied node when searching route
• [X] Robot keeps waiting if no new route available
• [x] Robot considers waiting-time cost and distance cost before using a new route
• [x] Multi-picker task allocating

[ZuyuanZhu]

• [x] Detect deadlocks , try to find and use a new route if there is one. But if not(e.g. two deadlocking robots are in a single track route or are surrounded by many other robots), a general solution needs to be done.

Todo:

• [ ] Create a general solution for deadlocks
• [ ] Split the current container-tmap out, for easy integration with other repositories. Topomap extended with simpy containers in topological_simpy and agents and picking process in rasberry_des.

Next stage:

• [ ] Picker agents can be assigned the picking tasks (can be defined as edge task - pick along a row; in one direction or both directions), which start at a node and ends at another node, instead of the primary and secondary head lanes
• [ ] Define all such tasks (like picking, transporting) in a config file, extend the simulation to any given map.
• [ ] One of the changes that @Iranaphor has done is the removal of the distinction between robot and picker agents. A non-ROS base version of coordinator may also be needed when we integrate these two efforts

Please correct me if I missed any important item to do @marc-hanheide @gpdas

[marc-hanheide]

This has been untouched for very long. I think functionality is still very useful in principle, but what's the plan @ZuyuanZhu ?

[ZuyuanZhu]

This has been untouched for very long. I think functionality is still very useful in principle, but what's the plan @ZuyuanZhu ?

The features have been moved to rasberry_des and used for topological optimisation which has been used in the two submited papers recently. I am working on the third paper, after this, I plan to encapsulate the simpy.container-based route planner as an API.