How chetankandya/ai-robotics works
This repository is a ROS AI-robotics course project. Its highlighted product workflow compares three reactive obstacle-avoidance approaches: fixed-window control, tabular reinforcement learning, and deep reinforcement learning. Sensors and ROS topics/services provide observations and motion commands; learning approaches also train policies. The tree additionally contains earlier occupancy-map and motion-planning work, a TST task executor, and PID tuning tools. Relationships for sampled code are shown where source demonstrates them; lab 4 pairings are documented in the README, while unsampled runtime wiring is omitted.
Main entry points: tst_executor.py TST service · q_explore.py Q-learning trainer · ls_to_occ.cpp Occupancy mapping · motion_planner_interface.h Planner interface
Components 22
Mapping and planning
- Occupancy mapping
catkin_ws/src/air_labs/air_lab2/src/ls_to_occ.cpp - Occupancy grid
catkin_ws/src/air_labs/air_lab2/include/air_lab2/occ.h - Motion planner
catkin_ws/src/air_labs/air_lab2/src/motion_planner.cpp - Planner interface
catkin_ws/src/air_labs/air_lab2/include/air_lab2/motion_planner_interface.h - State validity
catkin_ws/src/air_labs/air_lab2/include/air_lab2/state_validity_checker.h - Geometric fitting
catkin_ws/src/air_labs/air_lab2/include/air_lab2/Mathematics/GteApprQuery.h
Task execution
- TST service
catkin_ws/src/air_labs/air_lab3/src/tst_executor.py - Drive executor
catkin_ws/src/air_labs/air_lab3/src/drive_to_executor.py - Explore executor
catkin_ws/src/air_labs/air_lab3/src/explore_executor.py - Record executor
catkin_ws/src/air_labs/air_lab3/src/record_executor.py - Semantic recorder
catkin_ws/src/air_labs/air_lab4/src/record_semantic_executor.py
Obstacle avoidance
- Fixed-window policy
catkin_ws/src/air_labs/air_lab4/src/fixed_window.py - Fixed-window navigation
catkin_ws/src/air_labs/air_lab4/src/move_to_goal.py - Robot controller
catkin_ws/src/air_labs/air_lab4/src/controller.py
Policy learning
- Q-learning policy
catkin_ws/src/air_labs/air_lab4/src/QLearn.py - Q-learning trainer
catkin_ws/src/air_labs/air_lab4/src/q_explore.py - Q-table navigation
catkin_ws/src/air_labs/air_lab4/src/move_to_goal_ql.py - Deep Q-network
catkin_ws/src/air_labs/air_lab4/src/dqn.py - DQN trainer
catkin_ws/src/air_labs/air_lab4/src/dqn_learn.py
Controller tuning
- PID autotuner
catkin_ws/src/air_labs/air_lab1/src/autotune.py - Controller evaluator
catkin_ws/src/air_labs/air_lab1/src/evaluate_controller.py
Other components
- Robot and sensors
Connections 18
- Robot and sensors → Occupancy mapping: sends scans
catkin_ws/src/air_labs/air_lab2/src/ls_to_occ.cpp - Occupancy mapping → Occupancy grid: updates grid
catkin_ws/src/air_labs/air_lab2/src/ls_to_occ.cpp - Motion planner → Planner interface: requests paths
catkin_ws/src/air_labs/air_lab2/src/motion_planner.cpp - Planner interface → State validity: checks states
catkin_ws/src/air_labs/air_lab2/include/air_lab2/motion_planner_interface.h - TST service → Drive executor: registers executor
catkin_ws/src/air_labs/air_lab3/src/tst_executor.py - TST service → Explore executor: registers executor
catkin_ws/src/air_labs/air_lab3/src/tst_executor.py - TST service → Record executor: registers executor
catkin_ws/src/air_labs/air_lab3/src/tst_executor.py - Drive executor → Robot and sensors: publishes goals
catkin_ws/src/air_labs/air_lab3/src/drive_to_executor.py - Explore executor → Robot and sensors: publishes waypoints
catkin_ws/src/air_labs/air_lab3/src/explore_executor.py - Fixed-window navigation → Fixed-window policy: uses approach
- Q-learning trainer → Q-learning policy: trains policy
catkin_ws/src/air_labs/air_lab4/src/q_explore.py - Q-table navigation → Q-learning policy: uses Q table
- DQN trainer → Deep Q-network: trains network
catkin_ws/src/air_labs/air_lab4/src/dqn_learn.py - Q-learning trainer → Robot and sensors: publishes actions
catkin_ws/src/air_labs/air_lab4/src/q_explore.py - Robot controller → Robot and sensors: controls motion
- Semantic recorder → Robot and sensors: records semantics
- PID autotuner → Controller evaluator: reads statistics
catkin_ws/src/air_labs/air_lab1/src/autotune.py - Controller evaluator → Robot and sensors: measures motion error
catkin_ws/src/air_labs/air_lab1/src/evaluate_controller.py