Active SLAM 主动SLAM

本博客仅为作者记笔记之用，不对之处，望见谅，欢迎批评指正。

更多相关博客请查阅：http://blog.csdn.net/weixin_39779106

如需转载，请附上本文链接：http://blog.csdn.net/weixin_39779106/article/details/78728527

1. 为什么要用Active-SLAM：So far we described SLAM as an estimation problem that is carried out passively by the robot, i.e. the robot performs SLAM given the sensor data, but without acting deliberately to collect it.
2. 什么是Active-SLAM：  controlling robot’s motion in order to minimize the uncertainty of its map representation and localization is usually named active SLAM
3. 如何进行Active-SLAM： leverage robot’s motion to improve the mapping and localization results.

   ---

4. 首次提出这一名词： The first proposal and implementation of an active SLAM algorithm can be traced back to Feder [104,105] while the name was coined in [190]. However, active SLAM has its roots in ideas from artificial intelligence and robotic exploration that can be traced back to the early eighties (c.f. [19]).
5. 重要论述：Thrun in [300] and [297] concluded that solving the exploration-exploitation dilemma, i.e., finding a balance between visiting new places (exploration) and reducing the uncertainty by re-visiting known areas (exploitation), provides a more efficient alternative with respect to random exploration or pure exploitation.
6. Active-SLAM的主要研究内容： 1）Theory of Optimal Experimental Design (TOED) which, applied to active SLAM， allows selecting future robot action based on the predicted map uncertainty. 2） Information theoretic in this case decision making is usually guided by the notion of information gain. 3) Control theoretic approaches for active SLAM include the use of Model Predictive Control A different body of works formulates active SLAM under the formalism of Partially Observably Markov Decision Process which in general is known to be computationally intractable; approximate but tractable solutions for active SLAM include Bayesian Optimization or efficient Gaussian beliefs propagation, among others.
7. 流行的Active-SLAM框架：Selecting the best future action among a finite set of alternatives.
   1) The robot identifies possible locations to explore or exploit, i.e. vantage locations, in its current estimate of the map; 机器人识别可用于探索或利用的地点，如现有地图中的有利地形。
   2) The robot computes the utility of visiting each vantage point and selects the action with the highest utility; 机器人计算访问每个有利位置的效用，并选择具有最高效用的动作
   3) The robot carries out the selected action and decides if it is necessary to continue or to terminate the task. 机器人执行选定的操作，并决定是否有必要继续或终止该任务。
   
   

---

On the Comparison of Uncertainty Criteria for  Active SLAM

1. 主要工作：

• 将最优实验设计理论用于Active-SLAM
• 证明D最优指标能够给出有效信息作为执行SLAM的机器人的不确定性的度量。

1. 好句：This approach is known as active SLAM and specifically refers to the problem of how to give a mobile robot the capability of generating on-line trajectories that simultaneously maximize the accuracy of the map and robot’s localization, regarding a SLAM task.

---

Adaptive Mobile Robot Navigation and Mapping  </p>