Scholarlink

<p>This paper investigates an olfactory-based navigation method that enables a Mars lander to target any time-varying methane plume during the powered descent, even if the source location is unknown a priori. The episodic methane plumes emanating from the Martian surface invite the possibility of direct access to the subsurface methane reservoir to acquire pristine organic material. Any descending lander adapted for plume localization is required to not only land safely, but it must also infer the location of the landing target - the plume source - autonomously during the descent. However, existing plume source localization methods, which are all discriminative model, are computationally complex and seriously overfitted, rendering them ineffective in Mars methane exploration missions. We propose a generative source localization method based on a novel Gaussian mixture plume model. (1) A Gaussian mixture model for continuous releasing plume the corresponding probabilistic graphical model are proposed. (2) A generative source localization method is developed. The proposed method first maximizes the posterior probability of plume capture events then calculate the source location by minimizing the residuals. The proposed method resolves the overfitting problem of the existing methods while reducing the computational complexity. (3) By integrating the proposed olfactory-based navigation method and the E-guidance law, a biomimetic powered descent navigation and guidance method for Mars methane exploration missions is developed. Simulations in turbulent environment developed via computational fluid dynamics determined that the proposed method could overcome the overfitting problem while reducing the computational complexity. Besides, the fuel consumption imposed by the proposed method is modest compared with the baseline scenario where the landing target is known a priori. We believe that the proposed method offers the prospect of targeting the methane vent sources for direct access to subsurface material prior to its ejection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Probabilistic generative olfactory-based navigation in turbulent environments for Mars methane exploration

  • 期刊论文

  • 发表日期：

    2025年2月7日

查看全文
  • Yue Sun,
  • Qingyuan Qi,
  • Yang Liu

摘要

This paper investigates an olfactory-based navigation method that enables a Mars lander to target any time-varying methane plume during the powered descent, even if the source location is unknown a priori. The episodic methane plumes emanating from the Martian surface invite the possibility of direct access to the subsurface methane reservoir to acquire pristine organic material. Any descending lander adapted for plume localization is required to not only land safely, but it must also infer the location of the landing target - the plume source - autonomously during the descent. However, existing plume source localization methods, which are all discriminative model, are computationally complex and seriously overfitted, rendering them ineffective in Mars methane exploration missions. We propose a generative source localization method based on a novel Gaussian mixture plume model. (1) A Gaussian mixture model for continuous releasing plume the corresponding probabilistic graphical model are proposed. (2) A generative source localization method is developed. The proposed method first maximizes the posterior probability of plume capture events then calculate the source location by minimizing the residuals. The proposed method resolves the overfitting problem of the existing methods while reducing the computational complexity. (3) By integrating the proposed olfactory-based navigation method and the E-guidance law, a biomimetic powered descent navigation and guidance method for Mars methane exploration missions is developed. Simulations in turbulent environment developed via computational fluid dynamics determined that the proposed method could overcome the overfitting problem while reducing the computational complexity. Besides, the fuel consumption imposed by the proposed method is modest compared with the baseline scenario where the landing target is known a priori. We believe that the proposed method offers the prospect of targeting the methane vent sources for direct access to subsurface material prior to its ejection.