Predicting Object Dynamics From Visual Images Through Active Sensing Experiences

An Approach for Robots to Deal with Objects

“…The concept of affordance has been coined by J.J. Gibson on the ecological approach to visual perception and its link to action. Although he introduced the concept in psychology, it turned out to be elusive concept that influenced studies ranging from ecology, art science, industrial design, human-computer interaction and robotics [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25].…”

“…Contradicting with the last two researchers above, they used low level features, which are extracted from stereo vision or range image and used in learning and predicting of traversability affordance in unknown environment. Different from previous researcher, in [24] and [25] used imitation learning algorithm in order a humanoid robot learns object's affordances. Lopez used a probabilistic graphical model known as Bayesian networks to encode the dependencies between actions, object features and the effects of those actions.…”

Detecting Features of Tools, Objects, and Actions from Effects in a Robot using Deep Learning

“…Despite this development process being known, most conventional research on robot tool-use has focused on either tools or objects. Nishide et al [4] and Montesano et al [5] enabled robots to learn the relations between actions and object movements. Along another line, Stoytchev [6], Nishide et al [7], Mar et al [8], and Takahashi et al [9] enabled robots to acquire tool function for a specific target object.…”