Roadmap 2019 (Draft)

Chapter 1 : Transition to version 2

1. General cleanup

• AI
  • Scheduler: Subscribe and wait for a message to have the right attributes until timeout
  • game_status: blocking arming step
• Navigator
  • Option to disable pathfinder
  • Respond "success" when asked to go where the robot already is (+- 5mm)
  • If start or end position is invalid, no need to retry
  • Moving close to an edge, or getting out when robot position is inside a wall impossible
  • Show in Rviz when no path found
• Collisions
  • Migration to C++
  • Selectable sources
• Sensors
  • Ability to disable layers and/or individual sensors everywhere (belt, lidar, collisions)
  • objects_classifier destroys object when inside the robot
• Actuators
  • Actuators/ax12 link
  • Startup sequence
  • Create generate_arduino.sh for roslib

2. Nice Rviz

• Panel for lidar_objects
• Obstacles markers displayed by obstacles_classifier, not collisions
• Nice robot STL
• Navigator displays the current navigation direction (forwards/backwards)
• Interactive markers ?
• TF visual tweaking (e.g. to easily setup the laser frame)
• Collisions highlights dangerous objects
• Display current path

3. Packages reorganization

• No more namespace_ in the package names
• Relative message paths
• ai_game_status + ai_timer = ai_game_game_manager
• New map strategy (static objects only, no more JSON)
• Extended objects_classifier to new recognizer node
• Nodes grouping into fewer packages

4. Development workflow

• Robots connect to a WiFi hotspot
• Test SCP workflow
• Cross-platform CI/CD compilation ?

Chapter 2 : A stable system

1. Precise navigation

Asserv

• Adjust PIDs (zero-or-full speed and no wobbling)
• Smart PWM™ ability + handle simulation
• ROS position back to asserv connection (option to SetPos only X, Y and/or A)

Localization

• IMU angle connection to localizer
  • Kalman filter on localizer

Displacement

• Implement displacement moments in scheduler
• Bump displacement method (Smart PWM™)
• LiDAR displacement method (wall, corner)

2. Intelligent navigation

Navigator & asserv

• Adjustable navigation speed
  • Scheduler option
  • Autonomous decision : if the travel is near an object, approaching a waypoint, short distance then navigate slowly
• Pathfinder modes : consider all objects, all except game objects, only the enemy, no pathfinder…
• Correct simulation when Smart PWM, bumping into a wall

Pathfinder

• Gives the turning direction for each waypoint
• Implement target_frame (+ offset) attributes for the navigation goal
• Automatic waypoint approach
• Backwards ability (after a collision & regular navigation)
• Speed management during collisions
• Pathfinder loginfo time statistics, logerr path not found…

3. Remade Sensors stack

Sensors & processing

• Simplified map : only holds static objects and waypoints, GET/SET not based on JSON anymore…
• objects_classifier generalized to objects_recognizer, outputs all vision objects from belt/LiDAR with labels (wall, robot, cube…)
  • Display labels in Rviz (+ change colors)
• belt_interpreter generalized to range_filler, supports 1D sensors (alone for displacement or arrays for the belt)

Collisions

• Options to ignore layers or sensors
• Feed obstacles from obstacles_classifier, not map
• Get static objects from map

4. Nice touches

• Beautiful Rviz
• HMI
• Button to restart the whole system
• LEDs and confirmation for 12V
• Team indication on all screens
• Connect everything directly to the RPi ?

5. Adaptive AI

• Implement task pauses and resumes (resume only at the end of the regular tree)
• Feed task responses into other tasks' requests
  • Conditions before task executing

6. Buddy communication

• Implement full communication between two robots (multi-frequency!)
• Feed the robots' positions to the collisions nodes.

Chapter 3 : Making it Badass

1. Navigation

• Put the robot (almost) anywhere on the table and it will find its position & move to the spawn

2. Enemy tracking

• Disable scheduler actions based on the enemy's position
• Predict whole enemy navigation paths (=> better collisions management)

3. Badass AI

• Simultaneous tasks (TODO needed?)

4. Displacement

• Navigation strategies for displacement corrections
• LiDAR displacement on corners

5. Vision

• Camera support
• Industrial displacement sensors