SR-GRAT:对称响应引导增强学习与适应准,用于敏捷的固定翼无人机控制
IEEE transactions on cybernetics
|February 11, 2026
概括
本研究引入了用于飞机自动驾驶控制的新增强化学习 (RL) 框架. 它使用自适应式学习曲线和动态目标来提高稳定性并加快复杂的空中任务的学习速度.
科学领域:
- 航空航天工程 航空航天工程
- 人工智能的人工智能
- 控制系统 控制系统
背景情况:
- 传统的自动驾驶控制方法通常使用静态学习进度.
- 这可能会导致低于最佳的性能,代理人被压倒或停滞不前.
- 在动态的空中环境中需要适应性控制策略.
研究的目的:
- 为固定翼飞机自动驾驶控制提供一个新的课程框架.
- 通过适应性和动态机制提高学习效率和稳定性.
- 在复杂的空中任务场景中证明框架的有效性.
主要方法:
- 引入了对称响应引导强化学习 (RL).
- 实施了基于奖励趋势调整的自适应双向学习曲线.
- 在剧集内使用了动态目标生成机制.
主要成果:
- 在轨迹跟踪任务中实现了更快的趋同,减少了超越.
- 证明了更准确的跟踪性能,即使在流下.
- 在途径点导航和动态追逐方面表现出优于基线方法的优势.
结论:
- 拟议的RL框架为飞机控制提供了更好的稳定性和更快的融合.
- 动态目标生成可以提高学习效率,而不需要手动塑造奖励.
- 该控制器是强大的,适用于复杂的任务,如自主空中战斗.
相关概念视频
Cells of the Adaptive Immune Response
9.1K
The T and B lymphocytes of the adaptive immune system develop from common lymphoid progenitor cells in the bone marrow. These progenitors give rise to precursors that eventually develop into both T and B lymphocytes. As these precursors mature, they gain the ability to detect and respond to foreign antigens in the body, a process known as immunocompetence. Additionally, these precursors acquire self-tolerance, a process that ensures they do not react to self-antigens. This intricate system...
9.1K
Conservation of Mass in Fixed, Nondeforming Control Volume
1.6K
The principle of conservation of mass is fundamental in fluid dynamics and is crucial for analyzing flow within fixed control volumes, such as pipes or ducts. This principle states that the total mass within a control volume remains constant unless altered by the inflow or outflow of mass through the control surfaces. This results in a vital relationship for steady, incompressible flow where the mass entering a system equals the mass leaving it.
In the case of a sewer pipe, which can be modeled...
In the case of a sewer pipe, which can be modeled...
1.6K
Symmetric Member in Bending
618
In the study of the mechanics of materials, analyzing the behavior of prismatic members under opposing couples is crucial for understanding internal stress distributions, which are essential for structural design. When subjected to couples, a prismatic member experiences internal forces that maintain equilibrium. A couple, characterized by two equal and opposite forces, creates a moment but no resultant force. The internal forces at any section cut of the member must balance these external...
618
Target Cell Response to Hormones
5.8K
Hormones intricately bind to receptors on the surface or within target cells, initiating a cascade of cellular responses.
Notably, the cellular response can be regulated by altering the number of receptors expressed in the cell. For example, prolonged exposure to elevated hormone levels results in a gradual decline or down-regulation in the number of receptors for that specific hormone on the cell surface. Conversely, in response to low hormone levels, cells may use up-regulation, producing an...
Notably, the cellular response can be regulated by altering the number of receptors expressed in the cell. For example, prolonged exposure to elevated hormone levels results in a gradual decline or down-regulation in the number of receptors for that specific hormone on the cell surface. Conversely, in response to low hormone levels, cells may use up-regulation, producing an...
5.8K
Fixed Action Patterns
17.7K
A fixed action pattern (FAP) is a specific, hard-wired sequence of behaviors that occurs in response to an external stimulus, called a sign stimulus. The behavior is “fixed” because it is essentially unchangeable—proceeding similarly across individuals of a species every time it occurs.
17.7K
Reinforcement
952
Positive and negative reinforcement are key concepts in operant conditioning, a learning process where the consequences of a behavior affect the likelihood of that behavior being repeated.
Positive reinforcement occurs when a behavior is followed by the presentation of a rewarding stimulus, increasing the frequency of that behavior. For example:
Positive reinforcement occurs when a behavior is followed by the presentation of a rewarding stimulus, increasing the frequency of that behavior. For example:
952


