A simulated TurtleBot autonomously navigates and maps an environment.

• Occupation Grid processing was done to aid in navigation.
  • Morphological operations to make sure the robot does not collide with obstacles.
  • Convolution smoothing (values used in A*) so that the robot prefers to keep distance from obstacles.
• BFS-based frontier exploration algorithm detects the closest frontier border at each time step.
• A* algorithm plans a path from the robot to the center of the frontier border, using processed Occupancy Grid values as additional cost in order to prioritize open spaces.
• PID controller guides the robot along the path.
• Additional checks for emergency situations, and subsequent recovery behaviour were also implemented.

A simulated Turtlebot must navigate to and classify images in an environment within an allotted time.

• Brute-force algorithm finds the optimal path for navigating to all images.
• Custom algorithm navigates to location where the image is within FOV at an appropriate angle and distance.
• SURF feature detector extracts features from a cropped view and compares with the image database.
• Image classification algorithm select which database image it is seeing.

A simulated TurtleBot must explore an unknown environment and find 7 victims, classify their emotions, and interact approproately.

• Open-source frontier exploration algorithm was modified for improved performance and reliability.
• CNN was trained to classify 10 emotions.
• K-Folds Cross Validation used to select the best CNN model.
• Motion, sound, and visual interaction is performed for each victim.