The minimum data and nodes for autonomous driving with a real car are described. Some typical options are also listed. You should select what you need, and should make risk assessment based on your purpose. More detailed descriptions are work in progress.

Mandatory

This mandatory components make a car run on a given way points. The behavior of the car is controlled only by the given data and the results of localization. It can not deal with external actions.

Data

• TF
• Point Cloud Map
• Way Points

Sensing

• LiDAR

Localization

• voxel_grid_filter
• ndt_matching

Connector

• vel_pose_connect

Planning - Mission

• waypoint_loader
• lane_rule
• lane_stop
• lane_select
• obstacle_avoid (not for obstacle avoid)
• velocity_set

Planning - Motion

• pure_pursuit
• twist_filter

Actuation

• different for each car type

Obstacle Stop

With this option, the car can stop short of obstacles on way points. If the obstacles are removed, the car starts running again.

Sensing

• ray_ground_filter (or ring_round_filter)

Real Road

With this option, the car can run on real road without traffic light. Cross walk, stop line, or etc, included in the data can be considered.

Data

• Vector Map

Traffic Light

With this option, the car can stop when the traffic light is red.

Data

• Vector Map

Sensing

• Camera

Traffic Light Recognizer

• feat_proj
• region_tlr

For Localization Feedback

Connector

• can2odom