公开(公告)号：US11900233B2

公开(公告)日：2024-02-13

申请号：US17826050

申请日：2022-05-26

申请人： Unity IPR ApS

发明人： Arthur William Juliani, Jr. ,  Mohamed Marwan A. Mattar

IPC分类号： G06N3/006 ,  A63F13/67 ,  G06N20/00 ,  G06F18/214 ,  G06V10/764 ,  G06N3/02

CPC分类号： G06N3/006 ,  A63F13/67 ,  G06F18/214 ,  G06N20/00 ,  G06V10/764 ,  G06N3/02

摘要： In example embodiments, a method of interactive imitation learning method is disclosed. An input is received from an input device. The input includes data describing a first set of example actions defining a behavior for a virtual character. Inverse reinforcement learning is used to estimate a reward function for the set of example actions. The reward function and the set of example actions is used as input to a reinforcement learning model to train a machine learning agent to mimic the behavior in a training environment. Based on a measure of failure of the training of the machine learning agent reaching a threshold, the training of the machine learning agent is paused to request a second set of example actions from the input device. The second set of example actions is used in addition to the first set of example actions to estimate a new reward function.

公开(公告)号：US11369879B2

公开(公告)日：2022-06-28

申请号：US16657868

申请日：2019-10-18

申请人： Unity IPR ApS

发明人： Arthur William Juliani, Jr. ,  Mohamed Marwan A. Mattar

IPC分类号： A63F13/67 ,  G06K9/62 ,  G06N20/00 ,  G06N3/02

摘要： In example embodiments, a method of interactive imitation learning method is disclosed. An input is received from an input device. The input includes data describing a first set of example actions defining a behavior for a virtual character. Inverse reinforcement learning is used to estimate a reward function for the set of example actions. The reward function and the set of example actions is used as input to a reinforcement learning model to train a machine learning agent to mimic the behavior in a training environment. Based on a measure of failure of the training of the machine learning agent reaching a threshold, the training of the machine learning agent is paused to request a second set of example actions from the input device. The second set of example actions is used in addition to the first set of example actions to estimate a new reward function.