What is the current behavior when a bicycle/motorcycle pass the ego vehicle from the side?

[VRichardJP]

It is very common for 2-wheel vehicles (bicycle/motorcycle) to drive on the side of the road and pass a column of cars (typically waiting for a traffic light), like in the picture below:

In that situation, if we imagine the dark brown car in front is running autoware, what is its behavior when it is passed by the cyclist?

It seems to me that if the ego vehicle is stopped (e.g. traffic light was red and just turned green), the surround obstacle checker will prevent the vehicle from driving until the cyclist is far enough. So if we imagine many other bicycles/motorcycles are following after, the ego vehicle may end up being stuck for a while. Is that assumption correct?

What about if the ego vehicle is still driving? Can the map based prediction node predict the correct trajectory for the cyclist (keep riding on the side, eventually following curved roads)? If so, would autoware keep its lateral position on the lane and drive forward? or generate an avoidance trajectory? or simply stop?

Does anyone has experience of this kind of scenario with a real vehicle?

[yukkysaito]

It seems to me that if the ego vehicle is stopped (e.g. traffic light was red and just turned green), the surround obstacle checker will prevent the vehicle from driving until the cyclist is far enough. So if we imagine many other bicycles/motorcycles are following after, the ego vehicle may end up being stuck for a while. Is that assumption correct?

I think it depends on the options user chooses.
Some users do not want the bus to start if there are obstacles nearby, while others want it to start in the situation described above.
I think it is good to have a choice of options depending on the user wants.
The current surround obstacle checker is a requirement for bus starting, so it may not be good for it as the above scene.

What about if the ego vehicle is still driving? Can the map based prediction node predict the correct trajectory for the cyclist (keep riding on the side, eventually following curved roads)? If so, would autoware keep its lateral position on the lane and drive forward? or generate an avoidance trajectory? or simply stop?

Regarding the current implementation:
I think it will follow or stop when it enters into the trajectory. (I believe there are options to select the following target.)
If it is not in the trajectory, it is ignored.

[VRichardJP]

Regarding the current implementation: I think it will follow or stop when it enters into the trajectory. (I believe there are options to select the following target.) If it is not in the trajectory, it is ignored.

What about the trajectory prediction for the cyclist? Does current implementation preserve longitudinal position of the obstacle? i.e. predicts the cyclist will stay on the left side of the lane (so maybe always stay outside the ego vehicle path); or does it predicts the cyclist will eventually reach the center line of the lane (in which case there would obviously be a collision?)