MengGuo / omni_fault

Motion and Task Planning for Omni-Directional Ground Vehicles

Single vehicle test

The robot is assigned a task to infinitely often load object a (either at region la1 with orientation west or at region la2 with orientation east), then unload it (either at region ua1 with orientation south or at region ua2 with orientation north) and visit its base h1, and which is specified as the LTL formula:

one_la = '(la1 && w) || (la2 && e)'
one_ua = '(ua1 && s)|| (ua2 && n)'
Y1_task = '[] <> ((%s &&ts loada) && <> (%s && unloada)) && [] <> (h1 && e)' %(one_la, one_ua)

• Initialize robot model in robot_model_def.py

• Start robot node via

python omni_planner.py Y1

• Manually send confirmation for robot next activity:

rostopic pub -1 activity_done_Y1 omni_fault/confirmation -- '0' 'goto' '1'
rostopic pub -1 activity_done_Y1 omni_fault/confirmation -- '1' 'la1' '1'

• Manually send operation mode update (faults) to robot via

rostopic pub -1  status_Y1 omni_fault/status -- 'type-II'

Then once the robot the current motion or action is confirmed (you could do it manually like above), you will see that the discrete plan has been updated due to the cost update in the underlying transition system.

Collaboration between multiple vehicles

There are four robots in

• Start all four robots via (in different terminals)

python omni_planner.py Y1
python omni_planner.py Y2
python omni_planner.py Y3
python omni_planner.py Y4

• Manually set robot Y2 stops due to type-III fault

rostopic pub -1  status_Y2 omni_fault/status -- 'type-III'

Then you will see the request and reply messages exchanged among the robots to negotiate which robot will share the task for Y2. Once the other robots accomplish their current motion or action (you could do it manually), e.g.,

rostopic pub -1 activity_done_Y3 omni_fault/confirmation -- '0' 'goto' '1'
rostopic pub -1 activity_done_Y4 omni_fault/confirmation -- '0' 'goto' '1'

then the task of Y2 will be assigned to one robot, depending on the reply messages.

Scenario description

Four robots (Y1, Y2, Y3, Y4) start from two home bases (h1, h2):

• two of them are assigned to loard object a from any of the loading stations (la1, la2) with a certain orientation (w, e, n, s), and then unload a to the unloading station (ua) with a certain orientation, and then return to its base to charge. This procedure is repeated infinitely often.

• the other two are assigned to loard object b from any of the loading stations (lb1, lb2) with a certain orientation (w, e, n, s), and then unload b to the unloading station (ub) with a certain orientation, and then return to its base to charge. This procedure is repeated infinitely often.

During this process, robot Y1 will experience Type-I failure and robot Y4 will experience Type-II failure. They both recover automatically by adapting its lower-level controller and high-level plan. Then robot Y3 will stop due to Type-III failure and its task will be re-assigned to another robot, based on online robot-to-robot communication.