Draw a motion path: Click on the path below to guide, and then draw a line with the pencil tool about how you want the object to move. Click the first frame of the path layer to place the object at the beginning of the line you just drew, and click the second frame to place the object at the end of the line.

Sampling-based algorithm is the mainstream algorithm at present, which can avoid local extreme value problem and is very efficient in practical implementation. In addition, the algorithm of directly modifying and optimizing trajectory also has good performance.

The grid search algorithm directly divides the configuration space into grids according to a certain resolution, and in each grid, the robot can move to the adjacent grid (based on the four-connected or eight-connected criterion). As long as the starting point and target point are specified in free space, Dijkstar, A* and other graph search algorithms can be used to search and solve.

Generally, low-dimensional problems can be solved by grid-based algorithms. The configuration space is directly divided into grids according to a certain resolution, and then the grids are connected by four-connectivity and eight-connectivity criteria, and then the search is carried out on the grids.

For high-dimensional programming problems, we can't explicitly describe the configuration space, but directly dividing the grid will introduce a lot of calculation. Optimization algorithms such as artificial potential field work well in high-dimensional situations, but are easy to fall into local minima.

Baidu Encyclopedia-Fitness Path