Hi, For yielding the car stops at the roundabout even when there is no car in it. Can you elaborate how is the yielding logic implemented? I see that there are certain areas marked on the map on which the ego vehicle pays attention.
Super nice!!! The visualization was in rviz2 AutowareAi or AutowareAuto or another tool? The white lane is global path possible paths, or represents the car turning radius🤔? I noted the green line is followed but not the white line options, why?
Hey Marcus, this was from 2018 and at the time rviz (from ROS 1) was used. Through 2019 and to now the community and Apex.AI is able to use all ROS 2-based tooling (e.g. rviz2, rosbag2, rqt, ros2cli, etc.). I'm not sure what you mean by the white lane -- the gray lanes show the pathways / road that exist in the HD map along with the semantic information like lane and light positions. The lightly transparent green line is the desired route. You'll notice the waypoints are also visualized and always are placed dead-center of the desired route. The waypoints serve as an input to the planner module. The white line that sweeps across the screen (and originates from the host vehicle) is showing the output of the planning module which is used by the controller. The planner uses the waypoints, a look-ahead distance, and other inputs to determine and output a trajectory. The trajectory is taken by the controller to determine control commands that will then drive the vehicle along the path (while avoiding obstacles detected by the perception subsystems). There is great material in the Autoware.Auto project (autowarefoundation.gitlab.io/autoware.auto/AutowareAuto/avpdemo.html) that provides details about the architectures used for specific ODDs in autonomous driving.
@@LyleAtApexAI amazing answer Lyle thanks a lot. I have already done all AutowareAuto course it is super good framework, you are right! However to complete my master research this software does not address my needs: Switch local planners or global planners and evaluate them doing a Benchmark of path length, path smootheness, trajectory execution time, etc as the MoveIt framework provides in ROS1 for antrophomorphic robots. I am starting using now AutowareAI and I would like to know from you that already used it. If this framework offers this possibility of exchange planners through the runtime manager interface. Actually my doubt is a little more deep to post here. Please if you can help me in this regard could you answer this question in git: gitlab.com/ApexAI/autowareclass2020/-/issues/89 I am kind of frustrated because I have spent a lot of months with Navigation Stack in ROS 1 to create a simulation, but there is just TEB local planner that works with cars...also spent too much time with AutowareAuto, and I expected AutowareAuto would offer the runtime-manager interface, but they did not maybe because it is used for commercial purposes, who knows. And I ask myself if AutowareAI will offer this Benchmark plots (maybe through Rosbag) or at least to switch at least 3 different local planners or global planners. Would you know to answer this question my colleague? Thanks very much in advance
@@LyleAtApexAI Hi thanks very much I am doing my best! Actually I have installed Autoware.AI on last days and I am testing with SVL simulator! It is amazing. The only issue that I have is when creating the custom scenario that I need with Visual Scene Editor or Python API. And Aftewards integrating them with the Autopilot of AutowareAI- ROS bridge.... The SVL simulator open and closes after seconds (says cluster is busy, bridge was disconnected)...also it seems that ROS is not subscribing to the right topics! well is there a chanel where we could "exchange pictures" as we say here in Brazil... I mean I have a lot of videos and attempts that are almost working...the SVL staff also is trying to figure out the issues. However they just have a robust system for VSE/PythonAPI integration Apollo 5. I am thinking if I should learn Apollo5 is it ROS based too? Well my deadline to deliver master is too short, sorry to trigger Autoware staff everytime, but I really need a God hands! hehe Thanks Lyle!!!
Hi,
For yielding the car stops at the roundabout even when there is no car in it. Can you elaborate how is the yielding logic implemented? I see that there are certain areas marked on the map on which the ego vehicle pays attention.
Super nice!!! The visualization was in rviz2 AutowareAi or AutowareAuto or another tool? The white lane is global path possible paths, or represents the car turning radius🤔? I noted the green line is followed but not the white line options, why?
Hey Marcus, this was from 2018 and at the time rviz (from ROS 1) was used. Through 2019 and to now the community and Apex.AI is able to use all ROS 2-based tooling (e.g. rviz2, rosbag2, rqt, ros2cli, etc.).
I'm not sure what you mean by the white lane -- the gray lanes show the pathways / road that exist in the HD map along with the semantic information like lane and light positions.
The lightly transparent green line is the desired route. You'll notice the waypoints are also visualized and always are placed dead-center of the desired route. The waypoints serve as an input to the planner module. The white line that sweeps across the screen (and originates from the host vehicle) is showing the output of the planning module which is used by the controller. The planner uses the waypoints, a look-ahead distance, and other inputs to determine and output a trajectory. The trajectory is taken by the controller to determine control commands that will then drive the vehicle along the path (while avoiding obstacles detected by the perception subsystems).
There is great material in the Autoware.Auto project (autowarefoundation.gitlab.io/autoware.auto/AutowareAuto/avpdemo.html) that provides details about the architectures used for specific ODDs in autonomous driving.
@@LyleAtApexAI amazing answer Lyle thanks a lot. I have already done all AutowareAuto course it is super good framework, you are right! However to complete my master research this software does not address my needs:
Switch local planners or global planners and evaluate them doing a Benchmark of path length, path smootheness, trajectory execution time, etc as the MoveIt framework provides in ROS1 for antrophomorphic robots. I am starting using now AutowareAI and I would like to know from you that already used it. If this framework offers this possibility of exchange planners through the runtime manager interface. Actually my doubt is a little more deep to post here. Please if you can help me in this regard could you answer this question in git:
gitlab.com/ApexAI/autowareclass2020/-/issues/89
I am kind of frustrated because I have spent a lot of months with Navigation Stack in ROS 1 to create a simulation, but there is just TEB local planner that works with cars...also spent too much time with AutowareAuto, and I expected AutowareAuto would offer the runtime-manager interface, but they did not maybe because it is used for commercial purposes, who knows. And I ask myself if AutowareAI will offer this Benchmark plots (maybe through Rosbag) or at least to switch at least 3 different local planners or global planners. Would you know to answer this question my colleague? Thanks very much in advance
@@MarcusVinicius-lq3fe for sure, I'll do what I can to point you in the right direction in the GitLab issue.
@@LyleAtApexAI Hi thanks very much I am doing my best! Actually I have installed Autoware.AI on last days and I am testing with SVL simulator! It is amazing. The only issue that I have is when creating the custom scenario that I need with Visual Scene Editor or Python API. And Aftewards integrating them with the Autopilot of AutowareAI- ROS bridge....
The SVL simulator open and closes after seconds (says cluster is busy, bridge was disconnected)...also it seems that ROS is not subscribing to the right topics! well is there a chanel where we could "exchange pictures" as we say here in Brazil... I mean I have a lot of videos and attempts that are almost working...the SVL staff also is trying to figure out the issues. However they just have a robust system for VSE/PythonAPI integration Apollo 5. I am thinking if I should learn Apollo5 is it ROS based too? Well my deadline to deliver master is too short, sorry to trigger Autoware staff everytime, but I really need a God hands! hehe Thanks Lyle!!!
I did food delivery in your headquarter :) came here for check
Next time stay for a coffee and we can show you the car! We'll have to see about the check ;)