[Dense-Urban-ODD] Enhance Cruise Planner's Responsiveness to Other Road Users

[brkay54]

Checklist

• I've read the contribution guidelines.
• I've searched other issues and no duplicate issues were found.
• I've agreed with the maintainers that I can plan this task.

Description

Part of:

• [Autoware Labs] Dense Urban ODD - Reviewing and Tracking Failed Scenarios #7485

Dense Urban ODD Characteristics

• Low-Speed Driving: Typical speeds around 15 km/h.
• Complex Traffic Elements: Includes dense traffic with various dynamic interactions.
• Dynamic Agents: Frequent encounters with moving pedestrians and cyclists.
• Bidirectional Movement: Vehicles operate in lanes supporting bidirectional flow.
• Traffic Rule Non-compliance: Other road users may exhibit behaviors such as:
  • Driving in the wrong direction.
  • Cutting in unexpectedly from off-road areas.
  • Jaywalking pedestrians.
  • Improper overtaking by drivers.

In the context of Dense Urban ODD, we need to fine-tune the existing features of the obstacle_cruise_planner and potentially develop new features in speed planning to achieve the characteristics mentioned above.

Purpose

Improvement in Cruising for Cut-In and Cut-Out Obstacles

• The obstacle_cruise_planner package currently uses predicted objects in a limited way for speed planning. We need to consider objects further away from the trajectory by expanding the coverage area.
• The planner should respond more effectively to obstacles. If the predicted paths of other vehicles intersect with our planned path, Autoware should react much earlier to avoid potential collisions.

Late Reaction to Potential Collisions

cut-in-vehicle.mp4

Late Reaction for Pedestrian

late-response-pedestrian.mp4

• For objects that briefly enter and exit our path without posing a significant collision risk, the planner should avoid unnecessary reactions. This selective response will prevent abrupt and unnecessary slowdowns, ensuring smoother and more efficient driving.

Unnecessary Stop for Transient Object

stop-for-transient-object.mp4

Stop for a Vehicle Coming Fast from Behind

• This is a common scenario in our real vehicle testing. While the Ego is starting to drive, a vehicle approaches quickly from behind and overtakes the Ego. In this situation, obstacle_cruise_planner reacts uncomfortably. We want to make the obstacle_cruise_planner more reactive to vehicles approaching from behind.

Possible Approaches

The obstacle_cruise_planner is capable of handling the situations mentioned above. My primary goal will be to fine-tune the parameters of the package. If any function is inadequate in handling these situations, I will improve it.

Definition of Done

• Create a proposal for changing parameters.
• If the package is inadequate to handle scenarios, propose improvements.
• Ensure all scenarios for Dense Urban ODD pass.
• Create additional scenarios if necessary.