Please enable JavaScript.

Coggle requires JavaScript to display documents.

Research Line #masearch - Coggle Diagram

- - - - Information
        
        Non-adaptive:check:
        
        Adaptive:check:
      - Number of agent
        
        Single-agent:check:
        
        Multi-agent
        
        Voronoi Partition & Re-partition
        
        Global comms
        
        Local comms (Ongoing)
        
        Global GP
        
        Same graph :check:
        
        Different graph
        
        Individual GP
        
        With communication (limited by times/distance) (Ongoing)
        
        Without communication
      - Pursuit-evasion
        
        OpenAI particle env based
        
        Implemented algorithms
        
        MADDPG :check:
        
        MAAC (attention critic) :check:
        
        Adversarial training :check:
        
        Exploration based
        
        Pursuit-evasion in graph
      - Baselines
        
        Branch & Bound (not suitable)
        
        CMA-ES:check:
        
        CMA-ES with receding horizon:check:
        
        RAOr (TSP based):check:
        
        Ergodic search (not suitable)
        
        RRT* - Adaptive:check:
      - Viewpoint selection, hierarchical control
      - Dynamic environment
        
        Persistent monitoring
        
        Learnable number of GP observations :check:
        
        GP temporal prediction
        
        Spatio-temporal forecasting (Ongoing)
        
        Attention-based measurement selection
      - IPP on real robot experiment
      - Multi-resolution with Kalman update :check:
    - - Observe at vertices
      - Fixed frequency
- - - - Single-agent (Ongoing)
      - Multi-agent
    - - Implement our algorithm at local scale under TARE :check:
      - Global planner
- - - - decomposition for multi agent with different speed under multiple GPs
        
        ergodic search method
        
        cross-entropy method
      - Decomposition for multi agent with different sensing accuracy under a single GP
        
        cross-entropy method
        
        ergodic search method
    - - Search under the decomposed responsible map
        
        Ergodic search
        
        cross-entropy ergodic coverage adaptive velocity
        
        Ergodic search metrics comparison
        
        constant velocity and sensor
        
        different velocity
        
        different sensor
        
        cross-entropy ergodic coverage fixed velocity range
        
        Informative path planning