(Information Science Expert) Lecture on Cognitive Robotics
- Predictive learning through Autoencoders
- Object grasping task
- At a given moment, convert ①visual image, ②sound, and ③robot’s joint angles into low-dimensional Features using Autoencoders
- Combine these features and treat them as the values of a moment in Predictive learning
- By applying predictive learning, the object grasping task can be achieved
- When using Seq2Seq,
- Encode natural language using Predictive learning methods