codedao360 / drivingswarm

All things driving swarm with global issue tracker and goals.

Driving Swarm

All things driving swarm with global issue tracker and goals (and rather comprehensive wiki)

Goals

• Research swarm-robotic behaviour with robots capable of mapping and planning
• Swarm Applications
  • Foraging
  • Collective Search
  • Dispersion/Formation
• Robotic Topics
  • World Representation in the Context of Swarm Robotics
  • Planning agents within swarms (Cooperative planning, emergent effects in plans and plan execution)
  • Localisation in Robotic Swarms
• Swarm Topics
  • Integration of long term behaviour and short term planning
  • Integration of SI algorithms in robotics applications
  • Swarms with diversity in behaviour

Milestones

• Basic Architecture (wip)

  • Multi-Master ROS
  • ROS-Node controlling the behaviour (BehaviourAPI)
  • Nav-Stack for robot navigation
  • Basic behaviour with random movement and collision avoidance
  • Central point of control for the swarm (starting and stopping the robots, task distribution)
  • Optional Goal- Integrate Localisation

• Simulation (currently working with Gazebo, slow for many robots)

  • Multiple robots in the same environment
  • Possibly not with multi-master ros but different prefixes per robot conrtoller
  • Same codebase as the real robots

• Unified World Representation

  • Have one world reresentation that works for the swarm
  • Semantic Maps, Local Dynamic Maps, ...
  • Needs to include - actions are based on representation not on physical sensors, distinguishes between static environment, dynamic environment and other robots/actors
  • Stretch-Goal - representation contains virtual objects (i.e. pheromones, semantic information)
  • Gossiping protocols for spacial information

• Collective Path Planning

  • Solve movement conflicts by collectively agreeing on a trajectory for each robot in a neighbourhood
  • Different Strategies for solving conflicts based on neighbourhood size and interconnectedness of the swarm
  • See how different planners can deal with one scenario

• Localisation

  • Multilateration
  • Camera Tracking
  • Sensor fusion
  • AMCL with MAP (localisation part of SLAM)

• Full 30 robot swarm

  • Buy and assemble another 27 robots
  • currently 3 (plus 10)

• Implement Collective Behaviour

  • Formation Behaviours
  • Foraging (Indian Cross)
  • Collective Search (e.g. Minimize Distance to Cable Car)

Repository Structure

• gitman should now be used to check out submodules (cf. wiki)