A piece from Rutgers TV (RUTV) about the lab that included a live learning demo and several robots (Aibo, Mindstorms) and an interview with Michael Littman.
The Movement is indeed the most interesting thing. The kinematics is the beauty!!!! Decision making really cool as in Life....we explore to learn a policy and then exploit the best policy. Thanks for sharing this videos. I really love to know RL deeply. Experience is what I seek. Thanks!!!!
So what is so different about this? isn't it still a limited number of if-then-elses that go through a infinited number of loops until the goal is met?
The Movement is indeed the most interesting thing. The kinematics is the beauty!!!! Decision making really cool as in Life....we explore to learn a policy and then exploit the best policy. Thanks for sharing this videos. I really love to know RL deeply. Experience is what I seek. Thanks!!!!
amightyo 2 years ago
One possibility is to train the robot as a dog.
You give the robot a reward when a goal is done.
The robot will then increase the probability of choosing the used actions based on the obtained reward.
You could google for Reinforcement Learning to learn more.
fidlej 2 years ago
And what do the lines of code look like that determines if the action is working the best for the situation at hand or a new approach should be made?
So, keeping track of what worked, when it worked and how oftern it worked to form the path of your if then elses?
LuisureSuitLarry 2 years ago
One difference is that you could use memory to remember which action works the best in different situations.
fidlej 2 years ago
So what is so different about this? isn't it still a limited number of if-then-elses that go through a infinited number of loops until the goal is met?
LuisureSuitLarry 2 years ago