Framework

Class Diagram

classDiagram
direction LR
class Robot {
    +Array state
        +transition()
}
class Sensor {
    +sense()
}
class Task {
    +Robot robot
    +Sensor sensor
    +Array target_map
    +Array obstacle_map
    +float dt
    +step()
}
class Model {
    +learn()
    +predict()
}
class Objective {
    +eval()
}
class Planner {
    +plan()
}
class Controller {
    +control()
}

Robot --* Task
Sensor --* Robot
Model --* Task

Model --> Objective
Objective --> Planner
Planner --> Controller
Controller --> Task

Roadmap

Models

• Gaussian Process Regression (GPR): continuous
• Occupanocy Grid Map (OGM): discrete

Objectives

• Entropy

Planners

• Myopic Planner: produces a single informative waypoint
• Monte Carlo Tree Search (MCTS): sampling-based, producesa sequence of waypoints or actions
• Rapidly-Exploring Random Trees (RRT): sampling-based, producesa a sequence of waypoints
• Bayesian Optimization (BO): trajectory optimization, producesa a parameterized trajectory
• Covariance Matrix Adaptation Evolution Strategy (CMA-ES): trajectory optimization, produces a parameterized trajectory

Controllers

• Proportional Derivative (PD) Controller
• Model Predictive Controller (MPC)

Sensors

• Single-Beam Ranger: sonar
• Multi-Beam Ranger: LiDAR, depth camera

Robots

• Omni-Directional Robot
• Differential-Drive Robot
• Ackermann-Steering Robot

Tasks

Note

Navigation is not a RIG task.

• Navigation:
• Active Mapping
  • Environmental Mapping
  • Occupancy Mapping
  • Autonomous Inspection
  • Search and Rescue
• Active Localization
• Active Simultaneous Localization and Mapping (SLAM)

Task	Goal	Requires Target Map	Requires Obstacle Map	Considers Collision	Localization Algorithm	Mapping Algorithm
Navigation	Move from start position to goal position	❌	✅ Known	✅	❌	❌
Environmental Mapping	Build the target map	✅ Unknown	❌/✅ Free/Known	❌/✅	❌	✅
Occupancy Mapping	Build the obstacle map	❌	✅ Unknown	✅	❌	✅
Autonomous Inspection	Build the target map	✅ Unknown Dense	✅ Unknown	✅	❌	✅
Search and Rescue	Locate as many victims as possible	✅ Unknown Sparse	✅ Unknown	✅	❌	✅
Active Localization	Actively reduce localization uncertainty while navigation	❌	✅ Known	✅	✅	❌
Active SLAM	Actively reduce localization and mapping uncertainty	❌	✅ Unknown	✅	✅	✅