Introduction to robotics (for Cognitive Science)1 hour lecture + 2 hour exercises at computer
1. Action. Actuators. The robot gear control: via position, via speed. The robot arm control. Direct and inverse kinematics.
Exercise: iCubSim touching ball on table video 1
2. Perception. Sensors. The basic processing of sensor data: distance, camera image, depth map, tactile map.
Exercise: iCubSim seeing ball on table via color detection video 2
3. Control. Decomposition of the control system by function and by activity. Significance of real time. Sensory-motor approach to perception.
Exercise: simulated ALLEN exploring room video 3
4. The regular objects recognition. Hough transform. audio
Exercise: Boebot following ping-pong ball video 4
5. The irregular objects recognition. Feature detectors: phase correlation, SIFT/SURF/ORB.
Representation of objects in robot memory. Tracking. audio
Exercise: Boebot following selected object via phase correlation and via ORB. MIL tracker video 5
6. The object type recognition: Haar, LBPH, DOT/HOG. Cascade classifiers and regressors trained by gradient boosting.
Exercise: Boebot following face. iCubSim watching on faces and bodies. video 6
7. Cognitive approach to robot control. GOFAI, planning. STRIPS. Sussman anomaly. Frame problem.
Exercise: STRIPS and simulated SHAKEY video 7
8. Post-cognitive approach to robot control. Emergence of control in modular control architecture. Brooks' subsumption architecture. Embodiment. Situated robots.
Behavioral robotics. audio
Exercise: Simulated ALLEN combining activities. video 8
9. Minsky' society model of mind. Inspiration from Piaget's developmental psychology. Imitation performed by Robot.
Exercise: iCubSim interacting with human video 9
10. Robots recognizing and emulating emotions.
Example: Courting between robots according Freud's pleasure pump theory.
Robot KISMET. (Perceptron and autoencoder, deep learning for emotion recognition).
Exercise: iCubSim mimics emotions. video 10
11. Cloud technology for robots. Robot Pepper. Google Cloud. IBM Watson. MicroSoft Azure.
Exercise: Voice control of iCubSim via calling Google Cloud. video 11
12. Dennet's kinds of minds: Darwinian, Skinnerian, Popperian, Gregorian.
Exercise: iCubSim robots with various kinds of mind. video 12
- iCubSim which recognizes ball on table (one position in the exact front of the robot), takes the ball from the table, take it up and release (Rohil)
- iCubSim which recognizes positions of ball and hand. Randomly select joint and direction of move to put hand on ball by trial movements following feedback from the observed positions
- shape detection of white ball: use circle Hough transform and select only white object
- make photo of iCubSim head, print it out and rotate the picture in front of camera. Let iCubSim moves its head to the left or to the right side according rotation of the image. Employs ORB feature detector
- Control simulated ALLEN by voice (use Recognize4PC, e.g. call Google could for speech recognition)
- Play a suitable audioclip according emotion of person in the front of camera. Employ model created by deep learning.
- STRIPS controling simulated ALLEN to navigate to more places inside and outside of the room - iCubSim which mimics movement taken by camera in more sophisticated way