Feature/add multilevel crowdsim #174

tianenchong · 2020-12-29T02:52:37Z

Added a lean (minimum working version to avoid performance degradation) crowdsim map for office (1 level) and clinic (2 levels)

Merge prerequisite: multilevel crowdsim traffic editor pr

…unch.xml to office_crowd.launch.xml and added building_crowdsim build task for office world in CMakeLists.txt

…om ignition fuel online models to src before building. Original MaleVisitorPhone model is removed from rmf_demo_assets src to reduce bulkiness.

…_demo_assets

…odified office_crowd demo building.yaml with new yaml format.

…eep it.

…ill be done through model_downloader

…rowdsim

…vel_crowdsim

…cation

…tion

luca-della-vedova

Thanks for the PR! Before I try it out had a bunch of high level comments

luca-della-vedova · 2021-01-11T02:05:35Z

demos/launch/office_crowd.launch.xml

@@ -0,0 +1,51 @@
+<?xml version='1.0' ?>


Can we avoid creating a whole new file and just having the crowd sim be a launch parameter? Similar to what is done with ignition through the use_ignition launch parameter.

I am not sure about creating a whole new file, as there could be some legacy old files left over due to merging my old version with current rmf_demos. use use_crowdsim launch parameter can't be used here because even if we remove the crowdsim env variables in launch xml file to disable the plugin, the plugin section in sdf can still prompt the activation of plugin, and not being able to find the env variables can crash the demo.

That sounds strange, we use the trick of "having a plugin in the world file but not in the path" to allow models to work with both ignition and gazebo so not finding a plugin should print an error but not crash the simulation, for example we do it for differential drive robots like the Caddy.
I could reproduce the crash when the menge environment variable is not present but it seems to be related to the plugin's code rather than gazebo / ignition itself.
Also from my understanding the environment variable is only for menge files, the plugin will still be loaded. What you could do is a check to make sure the environment variable exists / the files exist before trying to initialize menge to avoid raising the exception.

luca-della-vedova · 2021-01-11T02:05:52Z

demos/launch/simulation.launch.xml

@@ -42,10 +45,14 @@
        name="IGN_GAZEBO_SYSTEM_PLUGIN_PATH"
        value="$(var plugin_path)"/>

-     <env
+      <env


Nit, remove this change?

pls go ahead to remove it. Apparently, the automatic style formatting does not apply to xml.

luca-della-vedova · 2021-01-11T02:06:07Z

rmf_dashboard_resources/office/main.json

-{"dispensers": {"coke_dispenser": {"icons": {"coke_dispenser": ""}, "location": {"x": 16.85, "y": -4.8, "yaw": 0, "level_name": "L1"}}, "coke_ingestor": {"icons": {"coke_ingestor": "/dispensers/coke_ingestor/humanIcon.png"}, "location": {"x": 20.95, "y": -6.95, "yaw": 0, "level_name": "L1"}}}, "robots": {"tinyRobot": {"icons": {"tinyRobot": "/robots/tinyRobot/tinyRobot.png"}, "places": {"supplies": [], "tinyRobot2_charger": [], "coe": [], "tinyRobot1_charger": [], "hardware_2": ["coke_ingestor"], "cubicle_2": [], "pantry": ["coke_dispenser"], "station_1": [], "lounge": [], "cubicle_1": [], "hardware_1": [], "station_2": []}}, "deliveryRobot": {"icons": {"deliveryRobot": "/robots/deliveryRobot/deliveryRobot.png"}, "places": {}}}}
+{"dispensers": {"coke_ingestor": {"icons": {"coke_ingestor": "/dispensers/coke_ingestor/humanIcon.png"}, "location": {"x": 20.95, "y": -6.95, "yaw": 0, "level_name": "L1"}}, "coke_dispenser": {"icons": {"coke_dispenser": ""}, "location": {"x": 16.85, "y": -4.8, "yaw": 0, "level_name": "L1"}}}, "robots": {"deliveryRobot": {"icons": {"deliveryRobot": "/robots/deliveryRobot/deliveryRobot.png"}, "places": {}}, "tinyRobot": {"icons": {"tinyRobot": "/robots/tinyRobot/tinyRobot.png"}, "places": {"supplies": [], "tinyRobot2_charger": [], "coe": [], "tinyRobot1_charger": [], "hardware_2": ["coke_ingestor"], "cubicle_2": [], "pantry": ["coke_dispenser"], "station_1": [], "lounge": [], "cubicle_1": [], "hardware_1": [], "station_2": []}}}}


Is this change necessary or can it be reverted?

There should be no change in this file, as the changes are due to the old remnants.

rmf_demo_maps/maps/airport_terminal/airport_terminal.building.yaml

rmf_demo_maps/maps/office/log.css

luca-della-vedova · 2021-01-11T02:09:32Z

rmf_demo_maps/maps/office/office.building.yaml

+      - [31, 13, {motion_axis: [1, start], motion_degrees: [3, 90], motion_direction: [2, 1], name: [1, main_door], plugin: [1, normal], type: [1, double_hinged]}]
+      - [35, 37, {motion_axis: [1, start], motion_degrees: [3, 90], motion_direction: [2, 1], name: [1, coe_door], plugin: [1, normal], type: [1, hinged]}]
+      - [18, 17, {motion_axis: [1, start], motion_degrees: [3, 90], motion_direction: [2, 1], name: [1, hardware_door], plugin: [1, normal], type: [1, hinged]}]


This change also seems unnecessary?

this change is auto generated upon save in traffic editor, essentially just some random reordering, which is deemed unnecessary.

luca-della-vedova · 2021-01-11T09:43:28Z

A few high level comments, other than removing the changes in the few remaining files that still have a diff:

I tried to run the office demo and I noticed that sometimes the actors don't update their yaw and this results in the following behavior, did you notice this before?

I also noticed quite a few crashes at startup for ignition but it's been quite some time since I ran the office demo there so they might be unrelated.

Also apart from making humans configurable at runtime I'd probably recommend reducing the number of humans in the office environment and making the ones in the clinic level 1 move around as well.
Ideally they should be "a bit" of an obstacle for robots (hence not fully static like the ones in the clinic level 1) but not "too much" (hence not crowded like the office).
You can probably try to run a few robot loops to have an idea of how much delays the humans cause to the robots.

tianenchong · 2021-01-11T10:00:52Z

I did noticed the yaw glitch. I have not been paying much attention to the glitch as majority of the human are not suffering from it. I have yet to find out what causes it but I do suspect the glitch happen as a result of narrow lane width from the start. Btw, narrow lane width also plays apart in traffic congestion, blocking robot's way with human standing around it using mobile phone. I will make adjustments to both maps tomorrow.

… added new goalsets and states

…lt, where use_crowdsim:=1)

demos/launch/simulation.launch.xml

luca-della-vedova · 2021-01-13T07:21:53Z

I didn't foresee the issue that when the plugin is disabled the humans will be just walking in place. I'll update the model on fuel so they have an Idle animation instead but we should still make sure that the simulations work correctly when the humans are not spawned (specifically that humans are spawned far away from robot lanes so they will not be on the way of robots if they are not walking).

tianenchong · 2021-01-13T08:08:37Z

In office demo, when the crowdsim is disabled, all humans will overlap at origin (0,0,0), away from office at top left corner, so they will not get in the way of robot for this map. The humans only move into their designated position in sdf file when the gazebo clock starts. I am not sure if this is a quirk for actor/non-static model in gazebo. In other maps where the origin is somewhere in the building, this could be a problem. Nevertheless, all actors are overlapped into one.

tianenchong · 2021-01-13T08:11:54Z

Ignition and gazebo config resource path are different with "_ign" extension for the ignition map.

luca-della-vedova · 2021-01-13T08:25:49Z

In office demo, when the crowdsim is disabled, all humans will overlap at origin (0,0,0), away from office at top left corner, so they will not get in the way of robot for this map. The humans only move into their designated position in sdf file when the gazebo clock starts. I am not sure if this is a quirk for actor/non-static model in gazebo. In other maps where the origin is somewhere in the building, this could be a problem. Nevertheless, all actors are overlapped into one.

I noticed that gazebo and ignition have very different behavior here. In ignition the actor is in the right position but playing the wrong animation (walking animation in the correct spawn point) while in gazebo it is in the wrong position but with the right animation (idle, all in the origin).
The best of both worlds would be to have actors spawn in the right position with the idle animation in both simulators. This is a change for traffic_editor simulation plugins though, the only part to make sure in this PR is that the spawning point is not in the middle of a robot lane (as is the case in both the maps now).

Ignition and gazebo config resource path are different with "_ign" extension for the ignition map.

True I didn't notice, duplicated is good then.

luca-della-vedova · 2021-01-18T05:06:51Z

There have been quite a lot of changes since the dispatcher_demo PR was merged so these maps don't run anymore on the latest rmf_core.
Can you either rebase or merge the changes from the current main branch?

tianenchong · 2021-01-18T06:10:00Z

Sure.

tianenchong · 2021-01-18T06:20:50Z

I have resolved the clinic map differences but haven't tried it. For office map, unfortunately, the vertices have changed too much and hence affects their reference indices that I have to remake the map.

I can't merge it today because I still need the old map and old setup to figure out the origin of the yaw glitch.

…re/add_multilevel_crowdsim

luca-della-vedova · 2021-01-22T02:49:48Z

I have resolved the clinic map differences but haven't tried it. For office map, unfortunately, the vertices have changed too much and hence affects their reference indices that I have to remake the map.

I can't merge it today because I still need the old map and old setup to figure out the origin of the yaw glitch.

Any update on this? I am unable to test the demos since they don't run anymore with the latest rmf_core

luca-della-vedova · 2021-02-01T03:20:00Z

It seems that by default humans are spawned in robot lanes, hence if crowdsim were to be disabled it would not be possible to run demo scenarios (example screenshot for office demo attached, but the same is true for the clinic demo).

We should be very careful about not breaking any existing behavior since I expect most people would want to run the demos without crowd simulation.

tianenchong · 2021-02-01T03:55:07Z

For this demo and the clinic one, some part of human lane and robot lanes do overlap. Some human lanes (not all, because some part of the map can be quite narrow) vertices are positioned only slightly off the robot lane to avoid head-on obstruction.
If I understand correctly, what u meant is to spawn human at vertices of human lanes that are at completely different position from those of robot lanes?

luca-della-vedova · 2021-02-01T05:16:02Z

what u meant is to spawn human at vertices of human lanes that are at completely different position from those of robot lanes?

Correct.
What I mean is that if the crowdsim was to be disabled the humans would be stuck in the middle of traffic lanes and it would not be possible to run the demos, hence their spawn point should not be in the middle of a robot's path (the screenshot I attached is such behavior, with the human there not moving and the robot unable to proceed).

luca-della-vedova · 2021-02-25T04:51:06Z

Because of the upcoming migration would it be possible to split this into a PR with only the changes to the launch packages (that can be merged straight away) and a "future work" change for rmf_demo_maps?

tianenchong · 2021-02-25T05:40:43Z

I think that should be possible as the launch script has crowdsim plugin off by default. The demo maps can be treated as future work since there isn’t much add on, apart from crowdsim tag for each level. Edwin is working with crowdsim branch btw.

Chong Tian En added 18 commits October 16, 2020 17:34

Added Fred's office crowd sim demo

99d7141

Added MaleVisitorPhone model, renamed launch file from office_cdsm.la…

3d45724

…unch.xml to office_crowd.launch.xml and added building_crowdsim build task for office world in CMakeLists.txt

Added airport terminal crowdsim demo

ce05c27

Edited CMakeLists.txt to enable downloading MaleVisitorPhone model fr…

1adc1a3

…om ignition fuel online models to src before building. Original MaleVisitorPhone model is removed from rmf_demo_assets src to reduce bulkiness.

changed bsdtar to unzip (more commonly used) in CMakeLists.txt in rmf…

350373a

…_demo_assets

removed airport_terminal_crowd temporarily, will add it back later. M…

53b3ee0

…odified office_crowd demo building.yaml with new yaml format.

MaleVisitorPhone model will be downloaded automatically, no need to k…

86a0672

…eep it.

restore to original CMakeLists.txt. MaleVisitorPhone model download w…

04c5f70

…ill be done through model_downloader

Merge https://github.com/osrf/rmf_demos into feature/add_multilevel_c…

49ed3c9

…rowdsim

removed office and clinic map for testing

a663827

Merge https://github.com/osrf/rmf_demos

0e17a03

Added crowdsim to office map

c6e1abf

made some adjustment to maps

fcd49fb

lean crowdsim demo added

dae242b

Merge remote-tracking branch 'origin/master' into feature/add_multile…

db5a694

…vel_crowdsim

added level 2 crowdsim for clinic

1380539

increase capacity for locations so that actor won't get stuck at a lo…

60d68d1

…cation

increase human lane width to reduce the possibility of traffic conges…

8806647

…tion

luca-della-vedova self-requested a review January 11, 2021 01:58

luca-della-vedova suggested changes Jan 11, 2021

View reviewed changes

Chong Tian En added 3 commits January 11, 2021 11:39

removed office_crowd.launch.xml

ca562c9

removed generated unnecessary changes

dacd2e5

restore original airport terminal map

af729b3

Chong Tian En added 3 commits January 12, 2021 11:07

changed office map - reduced humans to 3, increased some lane widths,…

eb9bd98

… added new goalsets and states

modified clinic map to include 3 humans on both levels

0d3c8d0

added crowdsim enable/disable feature in launch.xml (enabled by defau…

216abe9

…lt, where use_crowdsim:=1)

luca-della-vedova reviewed Jan 13, 2021

View reviewed changes

demos/launch/simulation.launch.xml Outdated Show resolved Hide resolved

disable crowdsim by default

b044eee

luca-della-vedova mentioned this pull request Jan 13, 2021

Feature/multilevel crowdsim open-rmf/rmf_traffic_editor#279

Closed

Chong Tian En added 2 commits January 21, 2021 17:14

removed office and clinic map to perform a clean merge

d5def94

Merge branch 'master' of https://github.com/osrf/rmf_demos into featu…

4ca6cbd

…re/add_multilevel_crowdsim

added crowdsim for office and clinic maps

ff22f86

move spawned humans off robot lanes

63fa8f5

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature/add multilevel crowdsim #174

Feature/add multilevel crowdsim #174

tianenchong commented Dec 29, 2020 •

edited

Loading

luca-della-vedova left a comment

luca-della-vedova Jan 11, 2021

tianenchong Jan 11, 2021 •

edited

Loading

luca-della-vedova Jan 12, 2021 •

edited

Loading

luca-della-vedova Jan 11, 2021

tianenchong Jan 11, 2021

luca-della-vedova Jan 11, 2021

tianenchong Jan 11, 2021

luca-della-vedova Jan 11, 2021

tianenchong Jan 11, 2021

luca-della-vedova commented Jan 11, 2021

tianenchong commented Jan 11, 2021 •

edited

Loading

luca-della-vedova commented Jan 13, 2021

tianenchong commented Jan 13, 2021 •

edited

Loading

tianenchong commented Jan 13, 2021

luca-della-vedova commented Jan 13, 2021 •

edited

Loading

luca-della-vedova commented Jan 18, 2021

tianenchong commented Jan 18, 2021

tianenchong commented Jan 18, 2021 •

edited

Loading

luca-della-vedova commented Jan 22, 2021

luca-della-vedova commented Feb 1, 2021

tianenchong commented Feb 1, 2021

luca-della-vedova commented Feb 1, 2021 •

edited

Loading

luca-della-vedova commented Feb 25, 2021

tianenchong commented Feb 25, 2021

		{"dispensers": {"coke_dispenser": {"icons": {"coke_dispenser": ""}, "location": {"x": 16.85, "y": -4.8, "yaw": 0, "level_name": "L1"}}, "coke_ingestor": {"icons": {"coke_ingestor": "/dispensers/coke_ingestor/humanIcon.png"}, "location": {"x": 20.95, "y": -6.95, "yaw": 0, "level_name": "L1"}}}, "robots": {"tinyRobot": {"icons": {"tinyRobot": "/robots/tinyRobot/tinyRobot.png"}, "places": {"supplies": [], "tinyRobot2_charger": [], "coe": [], "tinyRobot1_charger": [], "hardware_2": ["coke_ingestor"], "cubicle_2": [], "pantry": ["coke_dispenser"], "station_1": [], "lounge": [], "cubicle_1": [], "hardware_1": [], "station_2": []}}, "deliveryRobot": {"icons": {"deliveryRobot": "/robots/deliveryRobot/deliveryRobot.png"}, "places": {}}}} No newline at end of file
		{"dispensers": {"coke_ingestor": {"icons": {"coke_ingestor": "/dispensers/coke_ingestor/humanIcon.png"}, "location": {"x": 20.95, "y": -6.95, "yaw": 0, "level_name": "L1"}}, "coke_dispenser": {"icons": {"coke_dispenser": ""}, "location": {"x": 16.85, "y": -4.8, "yaw": 0, "level_name": "L1"}}}, "robots": {"deliveryRobot": {"icons": {"deliveryRobot": "/robots/deliveryRobot/deliveryRobot.png"}, "places": {}}, "tinyRobot": {"icons": {"tinyRobot": "/robots/tinyRobot/tinyRobot.png"}, "places": {"supplies": [], "tinyRobot2_charger": [], "coe": [], "tinyRobot1_charger": [], "hardware_2": ["coke_ingestor"], "cubicle_2": [], "pantry": ["coke_dispenser"], "station_1": [], "lounge": [], "cubicle_1": [], "hardware_1": [], "station_2": []}}}}

Feature/add multilevel crowdsim #174

Are you sure you want to change the base?

Feature/add multilevel crowdsim #174

Conversation

tianenchong commented Dec 29, 2020 • edited Loading

luca-della-vedova left a comment

Choose a reason for hiding this comment

luca-della-vedova Jan 11, 2021

Choose a reason for hiding this comment

tianenchong Jan 11, 2021 • edited Loading

Choose a reason for hiding this comment

luca-della-vedova Jan 12, 2021 • edited Loading

Choose a reason for hiding this comment

luca-della-vedova Jan 11, 2021

Choose a reason for hiding this comment

tianenchong Jan 11, 2021

Choose a reason for hiding this comment

luca-della-vedova Jan 11, 2021

Choose a reason for hiding this comment

tianenchong Jan 11, 2021

Choose a reason for hiding this comment

luca-della-vedova Jan 11, 2021

Choose a reason for hiding this comment

tianenchong Jan 11, 2021

Choose a reason for hiding this comment

luca-della-vedova commented Jan 11, 2021

tianenchong commented Jan 11, 2021 • edited Loading

luca-della-vedova commented Jan 13, 2021

tianenchong commented Jan 13, 2021 • edited Loading

tianenchong commented Jan 13, 2021

luca-della-vedova commented Jan 13, 2021 • edited Loading

luca-della-vedova commented Jan 18, 2021

tianenchong commented Jan 18, 2021

tianenchong commented Jan 18, 2021 • edited Loading

luca-della-vedova commented Jan 22, 2021

luca-della-vedova commented Feb 1, 2021

tianenchong commented Feb 1, 2021

luca-della-vedova commented Feb 1, 2021 • edited Loading

luca-della-vedova commented Feb 25, 2021

tianenchong commented Feb 25, 2021

tianenchong commented Dec 29, 2020 •

edited

Loading

tianenchong Jan 11, 2021 •

edited

Loading

luca-della-vedova Jan 12, 2021 •

edited

Loading

tianenchong commented Jan 11, 2021 •

edited

Loading

tianenchong commented Jan 13, 2021 •

edited

Loading

luca-della-vedova commented Jan 13, 2021 •

edited

Loading

tianenchong commented Jan 18, 2021 •

edited

Loading

luca-della-vedova commented Feb 1, 2021 •

edited

Loading