Skip to content

Team TODO list

Jump to bottom
James Dunn edited this page Oct 18, 2017 · 3 revisions

20171018 Update

Waypoint updater:

  • Needs acceleration velocity target ramp up
  • Yaw of initial position decides whether to reverse the base waypoints or not
  • Determine the lookahead value from max velocity

TL detection/classification:

  • Alternate approach of traffic light detection (bounding boxes)
  • Additional classifier or CV method to determine TL state

Udacity suggests the following order

  1. Waypoint Updater Node (Partial): Complete a partial waypoint updater which subscribes to /base_waypoints and /current_pose and publishes to /final_waypoints.
  2. DBW Node: Once our waypoint updater is publishing /final_waypoints, the waypoint_follower node will start publishing messages to the/twist_cmd topic. At this point, we have everything needed to build the dbw_node. After completing this step, the car should drive in the simulator, ignoring the traffic lights.
  3. Traffic Light Detection: This can be split into 2 parts:
  • Detection: Detect the traffic light and its color from the /camera/image_raw. The topic /vehicle/traffic_lights contains the exact location and status of all traffic lights in simulator, so we can test our output.
  • Waypoint publishing: Once we have correctly identified the traffic light and determined its position, we can convert it to a waypoint index and publish it.
  1. Waypoint Updater (Full): We will use /traffic_waypoint to change the waypoint target velocities before publishing to /final_waypoints. Our car should stop at red traffic lights and continue moving when they are green. If the light is yellow, check distance to "stop line" and with current velocity determine if it makes sense to stop or continue through the intersection.

We need to break these major items into smaller tasks.

See also the repository Projects tab and current issues and milestones.

Updates

I've found that when I uncomment traffic lights in server.py (line 56), the message load increases to the point of non-utility (i.e. the car doesn't move.) So we need a way to reduce/filter the number of messages from that event emission.

We have the vehicle moving and steering (in the master branch). Yousof has made progress with the waypoint updater in his branch. A few more items are required: we need to add brakes and throttle control, detect a traffic light (perhaps easiest to use the position of the light in the array - e.g. if the bottom one is lit, it is likely green, the top one red.), and plan to stop ahead of the light. Note: The light changes to green and almost immediately back to yellow and then red again. Therefore we will need to ignore the light once the green event is detected and we start moving through the intersection.

If you have ideas on these areas or others, please post a note on Slack. We really need some focus on the traffic light section of the code.

Clone this wiki locally