面向火星车的复杂环境实时建模与自主行为优化

Usually the complicated environment is unstructured. For example, when the Mars rover moves on the Mars' terrain, it will encounter uncertainty problems brought about by the random distribution of the meteorite craters, or by the variations of the climate. Based on the constraints on the robot's mobile ability and the unknown environment, a robust adaptive path following method is proposed, aiming at improving the ability to deal with emergency situations. To tackle the dynamic environment modeling problem, an isomerism vision navigation system constituted by a stereo vision and an omni-directional vision is set up and an efficient simultaneous global localization and map building method is proposed. The online real-time environment modeling problem is solved by decreasing the complexity of the map building. Orientated at the incremental environment modeling, an autonomous navigation planning method is proposed by combining the "reactive activity planning" and the "cautious activity planning". To deal with the robot's adaptation inability facing dynamic environment, a humanoid action method is proposed to solid the theoretical basis for the mobile robot's real-time environment modeling and behavior optimization in unknown complicated environments such as the Mars.

复杂环境通常具有非结构特性, 如火星车在火星表面地形上的移动时, 陨石坑和沙坑分布情况及气候的多变性等因素，给火星车自主探测带来了不确定性。本项目结合移动机器人运动性能约束及相关未知环境属性，提出鲁棒自适应路径跟踪方法，提高处理突发情况的能力；针对环境特征的动态建模问题，建立基于立体视觉和全维视觉构成的异构多视觉导航系统，提出高效的同步全局重定位及地图重建理论、方法；通过研究降低地图创建复杂度方法，解决在线实时环境建模问题。面向非完整环境地图的增量式创建特点，采用"反应式行为规划"与"慎思式行为规划"相结合的复合式自主导航规划方法。针对现有移动机器人控制技术存在对未知环境动态适应性不足的问题，提出基于仿人动觉图式方法对移动机器人面向环境的自适应动态行为决策进行理论研究，从而为移动机器人在火星等未知复杂环境中的实时建模及自主行为优化提供了理论基础和技术支撑。

近年来，国际航天界又掀起了探索火星的热潮。各国都在实施各自的探索计划。对于火星车而言，最重要的能力是能够感知外界环境并根据外界环境改变自己的运动状态，而火星表面的复杂性给火星车自主探测带来了不确定性。本项目主要以实验室已有的月球车为参考设计并开发出了高折叠比的六轮独立驱动火星车原型机，建立了运动学与动力学模型，搭建了双目立体相机与全景相机构成的异构机器视觉系统，实现了火星车实时三维地图创建与同步定位，路径可通过性与障碍特征分析，建立了火星车运动规划算法以及越障、避障策略，引入词袋模型与图优化模型对火星车地图创建与定位进行了闭环检测，实现了火星车的自主导航与行为优化。本项目在webots仿真环境中完成了算法的验证，说明了本项目的成果对未来移动机器人在火星等未知复杂环境中的实时地图创建与同步定位及自主行为优化等问题提供了有效的理论基础和技术支撑。