The application of reinforcement learning is widely used by multi-agent systems in recent years. An agent uses a multi-agent system to cooperate with other agents to accomplish the given task, and one agent′s behavio...The application of reinforcement learning is widely used by multi-agent systems in recent years. An agent uses a multi-agent system to cooperate with other agents to accomplish the given task, and one agent′s behavior usually affects the others′ behaviors. In traditional reinforcement learning, one agent takes the others location, so it is difficult to consider the others′ behavior, which decreases the learning efficiency. This paper proposes multi-agent reinforcement learning with cooperation based on eligibility traces, i.e. one agent estimates the other agent′s behavior with the other agent′s eligibility traces. The results of this simulation prove the validity of the proposed learning method.展开更多
In the near future, active safety systems will take more control over the vehicle driving, even up to introducing fully autonomous vehicles. Nowadays, it is expected that the active safety systems will aid avoiding co...In the near future, active safety systems will take more control over the vehicle driving, even up to introducing fully autonomous vehicles. Nowadays, it is expected that the active safety systems will aid avoiding collisions much more efficiently than human drivers. These systems can protect not only the passengers, but also other road users. To mitigate collision, certain maneuvers （e.g., sudden braking, lane change, etc.） need to be done in a reasonably quick time. However, this may lead to low-g energy pulses. The latter fact, may cause unexpected and, in some cases, unwanted occupant body motion resulting even in OOP （out of position） postures. New patterns of occupant reactions in such cases are, to some extent, confirmed experimentally [1-3]. This paper evaluates the limits of standard ATDs （anthropometric test devices） and chosen human models in well established maneuver scenarios. Obtained results are compared with experimental data available in the literature. Drawbacks identify new challenges for the near future simulation based safety engineering. One scenario with combined conditions of emergency braking during lane change has been used as an example of OOP posture after maneuver.展开更多
文摘The application of reinforcement learning is widely used by multi-agent systems in recent years. An agent uses a multi-agent system to cooperate with other agents to accomplish the given task, and one agent′s behavior usually affects the others′ behaviors. In traditional reinforcement learning, one agent takes the others location, so it is difficult to consider the others′ behavior, which decreases the learning efficiency. This paper proposes multi-agent reinforcement learning with cooperation based on eligibility traces, i.e. one agent estimates the other agent′s behavior with the other agent′s eligibility traces. The results of this simulation prove the validity of the proposed learning method.
文摘In the near future, active safety systems will take more control over the vehicle driving, even up to introducing fully autonomous vehicles. Nowadays, it is expected that the active safety systems will aid avoiding collisions much more efficiently than human drivers. These systems can protect not only the passengers, but also other road users. To mitigate collision, certain maneuvers （e.g., sudden braking, lane change, etc.） need to be done in a reasonably quick time. However, this may lead to low-g energy pulses. The latter fact, may cause unexpected and, in some cases, unwanted occupant body motion resulting even in OOP （out of position） postures. New patterns of occupant reactions in such cases are, to some extent, confirmed experimentally [1-3]. This paper evaluates the limits of standard ATDs （anthropometric test devices） and chosen human models in well established maneuver scenarios. Obtained results are compared with experimental data available in the literature. Drawbacks identify new challenges for the near future simulation based safety engineering. One scenario with combined conditions of emergency braking during lane change has been used as an example of OOP posture after maneuver.
