起始页
Good afternoon, everyone. Today I’m going to introduce the application of the tree structure in path planning.
目录
The presentation will be divided into three parts : Overview, RRT algorithm, improved RRT algorithm.
Overview内容——1(机器人路径规划图)
As the name implies, path planning is to find a feasible path from the start position to the target position. With the rapid development of science and technology, more and more mobile robots are everywhere in people’s lives, and path planning is a crucial technology for mobile robots to successfully complete tasks.
Overview内容——2(3个图)
Here are some map structures commonly used in path planning.
The first is grid map. Each grid is in occupied state or idle state or unknown states.
The second is Octree Map structure.
Last is PointCloud.
Path planning by grid map in plane is used most.
Overview内容——3(四叉树规划)
Teacher Zhang mentioned quadtree in last class, which is used to divide space and then connecting blank block can get a path.
RRT算法——1(流程图)
RRT means Rapidly-exploring Random Tree. In RRT, the root node of the random tree presents the start point. Then leaf nodes are added through random sampling and collision detection to generate a random tree.
Here is a flow chart of RRT.
- Generate a random sampling point called NewNode, determine whether the point is in idle state. If not, resample NewNode.
- Traverse(遍历) the random tree and find the closest node to NewNode called ClosestNode.
- Check whether NewNode is within the step size limit of ClosestNode. If not, get a middle point to replace NewNode.
- Check whether there is an obstacle(障碍物) between NewNode and ClosestNode.If not, resample NewNode.
- If true, add NewNode to the random tree and set ClosestNode as the parent node of NewNode.
- Determine whether NewNode is within the step size limit of the end point.If true, perform collision detection on them. If true, the NewNode is set as the parent node of the end point, and the end point is added to the random tree to end the iteration. If not, continue iterating.
RRT算法——2(伪代码)
RRT算法——3(规画效果图)
- As shown in image(如图所示),the green point is the start point, and the yellow point is the end point.
- We can construct the random tree as KD-Tree to quickly search closet node.
改进的RRT算法
- Double trees RRT, one construct random tree with the start point as root node. and another with the end point as the root node. Util the two trees meet.
- RRT* obtains a set of adjacent nodes in the search tree within a circle whose radius is R and center is sampling node. Cut the nodes in the set and update the parent node of the node if there is a closer path to the start node.