1. Transfer Learning in Heterogeneous States and Tasks

This work only considers the same domain states in both the teacher and learner, and different domain tasks. For future research, we are interested in heterogeneous states and tasks for generalization purposes in robotics applications.

2. Imitation Learning

Until now, transfer learning is used for teaching to the chameleon-like robot the balancing task. However, the next step in a robot system is the walking task. In that sense, we are proposing to use imitation learning for extracting knowledge directly from a real chameleon and to use it in the chameleon-robot.