The Autoware Foundation
Autonomy Software WG Meeting 2022/03/08 #2470
Replies: 0 comments 5 replies
-
|
@xmfcx to create a Discussion for point cloud fields with the prev. mermaid diagram. |
Beta Was this translation helpful? Give feedback.
-
|
Today's presentation: Transition.to.Autoware.Core_Universe.pdf |
Beta Was this translation helpful? Give feedback.
-
|
Today's presentation. |
Beta Was this translation helpful? Give feedback.
-
|
@BonoloAWF Can the recording of this meeting be shared please? |
Beta Was this translation helpful? Give feedback.
-
|
Please ask @mitsudome-r for the recording. I don't have access. |
Beta Was this translation helpful? Give feedback.
-
Administrative
Attendees
Chaired by Fatih
Common Resources
Autoware.Auto
Autoware.Universe
Announcements
Discussion topics
Proposed by Mitsudome:
Mitsudome gave a presentation on the current status of the transition to Autowae.Core/Universe. See presentation slides in the comments below.
Summary of presentation:
Ambriose (A) - Will we be porting all the packages from Autoware.Auto to Universe? Mitsudome (M) - Maintainers will go through the Autoware.Auto packages to decide which should be moved to Universe. Fatih and I will create an initial list of candidate packages and ASWG members could help with porting to Universe. Fatih (F) - Universe is mainly the proposal from TierIV architecture. Let's first evaluate Universe and decide per package which to keep or replace based on performance.
Maxime - will the current Autoware.Auto be accessible? What about the docs rendering? M - We will archive Autoware.Auto and main development will move to Core/Universe. The documentation will remain the same and be located in the old Gitlab repo. It will be like Autoware.AI where no new features are being added.
Hatem (H) - the plan is to start developing bus ODD. Will it be in Core or Universe? How will previous ODDs (AVP, cargo delivery) be merged into Core? M - Core is currently empty. The proposed architecture for bus ODD will be in Universe then later refactored. After testing on a real vehicle, writing missing tests/docs to meet the coding requirements of Core, bus ODD packages will be ported from Universe to Core. H - will the node interfaces for other ODDs match with Autoware.Universe? M - the baseline autoware_auto_msgs will continue to be used. If we achieve the bus ODD features, we will have all the features in the AVP demo and cargo ODD demo. So people won't have to use both Autoware.Auto with Universe.
H - Autoware.AI allows having multiple nodes that can achieve the same objective (eg. 2 NDT nodes). So users could add different localisation packages for use. M - packages developed by the community in Universe can be merged to Core (subject to meeting the required standards). People can also use Core and select the Universe packages they want. So people could easily try out your contributed Universe packages. In Autoware.AI it was difficult to know which module replaces another and sometimes interfaces were not shared. We have now standardised the interfaces.
Timeline summary:
Fatih (F) - packages should have the same interface. We should finalise and standardise the msg infrastructure. Mitsudome (M) - yes, good point. F - next week we could do a demo of Universe architecture and later present the finalised message architecture. F - is universe working with SVL? M - I'm not sure if there is a connection to the SVL simulator. I need to check that since we are mostly using rosbags.
F - Hatem will you be able to test Autoware with Carla simulator? Hatem (H) - should I test with Universe repo or Autoware.Auto? Which is the highest priority? M - we are looking for a simulator for the bus ODD so test with Universe. We want to know how the data looks in ROS 2 and what the delays are. H - OK, I shall try to do that. F - I would like to see how it works with Autoware.AI too. H - I can can show at the next meeting how it works with Autoware.AI.
Proposed by Bonolo (Presentation):
Huei-Ru gave a presentation on the current state of the ITRI autonomous bus platform. She introduced the ITRI team that will be working on the bus ODD: Yi-Ling (perception) and Henry (decision control). See comments below for presentation slides.
Summary of presentation:
Fatih (F) - how many cameras and what lidars do you use? Huei-Ru (H) - 3x Ouster lidar and 8 cameras. Lidar is used for localisation. Mistudome (M) - do you drive backwards since there is no lidar at the back of the bus? H - we use cameras only at the back. M - so is the position estimate from the camera? What about obstacles behind the bus? You don't assume backward motion? H - correct.
F - what is the GPS/IMU? H - Trimble without RTK. F- what is the final output from the computers? Do you send acceleration, reference velocity to the vehicle? Henry - we send the gas paddle level for longitudinal control. For steering we send the steering angle and ROS topic to the steering CAN bus. For brakes we translate a voltage to brake pad level. F - what about the user interface? Henry - we monitor data through the lidar MPC including ROS topics.
Action Items
Review of issues
The last few remaining issues/MRs from the cargo ODD are expected to be completed this week.
Beta Was this translation helpful? Give feedback.
All reactions