innovator

Add idea


Calendar

«    January 2019    »
MonTueWedThuFriSatSun
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
 

 

Advert

 

Payment

 

Advert

 

Authorization

Стартап

DateDate: 2-01-2019, 09:11

The researchers said they developed an algorithm by which robots can learn to walk on their own. In a preprint document, scientists from the University of California, Berkeley and Google Brain describe a system that “taught a four-legged robot to cross the terrain — both familiar and unfamiliar.”
“Deep reinforcement learning can be used to automate robotic tasks, which allows for end-to-end learning that matches sensory data with low-level actions,” the authors explain. “If we manage to learn movements from scratch, we can make controllers that are perfectly adapted to each robot and even localities, potentially providing better maneuverability, energy efficiency and reliability.”
Strengthened learning is an artificial intelligence teaching technique that uses rewards or punishments to bring robots to the goal. It requires a large amount of data, in some cases tens of thousands of samples.
In experiments with OpenAI Gym, simulating an open source environment for teaching and testing AI agents, the authors model achieved “almost identical” or better performance compared to the baseline indicators for the four tasks of continuous movement.
Source: hightech.fm