针对的问题
- 鲁棒且精确的视觉定位，在 大场景 和 显著的场景外观变化 下仍有较大挑战
- 现有的运算资源开销太大，无法实时运行
提出的方法
提出了一个网络 HF-Net，能够预测图像的局部特征和全局描述子
定位过程分成两部：
- 首先使用全局描述子还原出候选场景
- 在候选场景中使用局部特征进行 6-DoF 位姿估计
实现的效果
- 节省运行时间，适用于实时运行
- 在场景有较大变化的情况下有较强的定位鲁棒性
[Paper Reading] From Coarse to Fine: Robust Hierarchical Localization at Large Scale
[Paper Reading] Autonomous Cave Surveying With an Aerial Robot (TRO2022)
针对的问题
- 全暗洞穴环境的无人机自主飞行
- 人工探险危险，因为存在洞穴结构不坚固
- 使用机器人建图可以降低危险，且需要实时回传信息，但现有的方法基本未考虑高分辨率感知建模的同时，降低高带宽，以在低带宽场景正常工作
采用的方法
- 传感器和硬件设备：
- 深度相机用于建图
- 朝下的相机用于状态估计
- 朝前和朝下的灯光照明
- this work compactly represents sensor observations as Gaussian mixture models and maintains a local occupancy grid map for a motion planner that greedily maximizes an information-theoretic objective function. The approach accommodates both limited field of view (FoV) depth cameras and larger FoV LiDAR sensors and is extensively evaluated in long duration simulations on an embedded PC.
达到的效果
无人机在洞穴自主飞行探索，实时回传三维地图
存在的不足/未来的工作
- 未对洞穴的一些特征（如钟乳石）进行语义分类和编码
- 3D 地图投影成 2D地图用于探险者使用，防止迷路
- 扩展成多机
- 引入场景重识别，以减少漂移，提高地图的一致性，也有利于多机飞行
P, NP, and NP-hard Completeness
What is P Class of Problem？
A problem has a polynominal algorithm means that it can be solved in polynominal time with respect to the size of inputs. We also call the problem as P problem.
Version of Problem
We have three version of problems. We consider problems where we have one “cost” that we want to minimize.
- Optimization: Given a problem instance, find the minimum cost solution.
- Evaluation: Given a problem instance, find the cost of the min-cost solution.
- Recognition: Given a problem instance and an integer L, is there a solution with cost no greater than L?
Note that the third version requires only a “yes-no” answer. This makes the discussion easier, and we will focus on Recognition problems. Depending on assumptions about the cost function, it can be shown that the three problem formulations are equivalent.
What is NP Class of Problem
Intuitively, this is the class of problems that given a problem instance and guess for a solution, we can verify in polynomial time, whether the guess is a correct solution.
Polynomial verification of a solution
For a reconization problem, if we are given a guess of solution, we want to verify whether this solution can help us answer this problem.
If we can “doublecheck” that the guess is a solution of the problem in polynomial time, we say that we can verify the problem in polynomial time.
Non-Deterministic Algorithm
Let’s imagine an algorithm for reconization problem, it somehow come up with a solution which may be generated by a guess, and check the cost of solution is less than L.
Therefore the algorithm can be decomposed in two parts: guessing and verifying
We call all problem instances for which answer is yes as yes-instance (There exists a solution of cost is less than L in a problem instance).
If a problem instance is a yes-instance, assuming we are good at guessing, we will find its solution with the first try, then we can solve the reconization in polynomial time:
\[guessing + poly.verification.\]However we can use this algorithm to solve the non-yes-instance (there is no solution with cost is less than L), because we have no way to prove that there is no such solution.
The NP Class
A problem belongs in Non-deterministic Polynomial class, if yes-solution of the problem can be verified in polynomial time.
WWFs (Well-Formed Fomulas)
There are eight WFFs: \(\begin{array}{ll} All \ A \ is \ B & x \ is \ A \\ no \ A \ is \ B \ & x \ is \ not \ A \\ some \ A \ is \ B & x \ is \ y \\ some \ A \ is \ not \ B & x \ is \ not \ y \end{array}\)
All the sentences not aforementioned are not WFFs.
referfence
Total views. cowboys. Hits