Skip to content

Latest commit

 

History

History
89 lines (76 loc) · 4.39 KB

OpponentTracker.md

File metadata and controls

89 lines (76 loc) · 4.39 KB
Raw

Obstacle Detection and Opponent Tracker

F1TENTH autonomous grand prix (ICRA'23 London) - ForzaETH

Launching in ICRA STACK

Launch headtohead with spliner:

roslaunch stack_master headtohead.launch LU_table:=NUCX_hangar_pacejka overtake_mode:=spliner od_mode:=sami

After successful launch, you should see the following messages in the command line:

  • [Tracking] Dynamic Tracker Server Launched...
  • [Perception Detect]: Ready
  • [Perception Classifier]: Ready

Dnyamic Reconfigurable Parameter Overview

Launch rqt and go into dynamic_tracker_server

Tracking

  • ttl_dynamic: the time to live of the dynamic prediction of the opponent.
  • ratio_to_glob_path: the ratio between the target velocity and the global velocity. Defines to which target speed the dynamic prediction will convert. If set to 1, target velocity is equal the raceline speed
  • ttl_static: time to live of the static and not-classified objects. Be aware that the ttl of static obstacle only gets decreased if they are/should be visible
  • min_nb_meas: minimum number of measurements of a obstacle before it gets classified as static or dynamic
  • dist_deletion: distance under which an obstacle can be deleted (if it is visible)
  • dist_infront: distance under which obstacles infront of the car get published
  • min_std: standard deviation of the position under which obstacles get classified as static. Needs to be increased when using pf2
  • max_std: standard deviation of the position over which obstacles get classified as dynamic. Needs to be increased when using pf2
  • vs_reset: average velocity in s direction under which a dynamic obstacle gets reset to a static obstacle
  • aggro_multi: a multiplier for the association distance for dynamic obstacles
  • debug_mode: will enable some prints to the console for debugging
  • publish_static: enable/disable publishing of static obstacles. May help when many FP are present
  • noMemoryMode: will disable the memory mode of the tracker. Can help with thight curves and trailing

Detect

  • min_obs_size: minimum size of the point could in laser points. Can be used to filter out FP
  • max_obs_size: maximum size of an object in meters. Can be used to filter out FP
  • max_viewing_distance: the maximum distance the detection can see. Maximum allowed for the race is 10m
  • boundaries_inflation: artificially increases the boundaries for detection. Can help to filter out FP along the boundaries.

Troubleshooting

Detection/Tracking not working

Did you build and sauce?

Trailing gets stuck in corners due to FP

  • Enable noMemoryMode
  • Increase boundaries_inflation

FP near the boundaries

  • Increase boundaries_inflation
  • Consider inflating the boundaries with map editor

Dynamic Obstalce is too fast/slow

  • change ratio_to_glob_path, 1 is raceline speed

Opponent/Obstacles are sometimes not visible

  • decrease min_obs_size
  • increase max_viewing_distance

Very big obstacles (FP)

  • decrease max_obs_size

Missclasification

  • increase min_nb_meas
  • increase min_std and max_std
  • incrase/decrease vs_reset

Obstacle is not deletet but clearly inside of visible range

  • enable debug_mode and fix it OR disable publish_static

Parameter in yaml

This variables should not need to be touched during the race. Listed for completeness.

Tracking

  • rate: set to 40Hz
  • P_vs: proportional gain for the s-velocity
  • P_d: proportional gain for bringing back the d-position to the center
  • P_vd: proportional gain for the d-velocity
  • measurment_var_s: the variance of the measurment noise in the s direction
  • measurment_var_d: the variance of the measurment noise in the d direction
  • measurment_var_vs: the variance of the measurment noise in the s-velocity
  • measurment_var_vd: the variance of the measurment noise in the d-velocity
  • process_var_vs: the variance of the process velocity noise in the s direction
  • process_var_vd: the variance of the process velocity noise in the d direction
  • max_dist: maximal distance for association of the obstacles
  • var_pub: obstacles with a bigger variance are not published

Detect

  • rate: set to 40Hz
  • lambda: minimum reliable detection range in degrees
  • sigma: standard deviation of the noise of the lidar ranges in meter
  • min_2_points_dist: minimum distance between two points

Go back to the checklist index