search

HanqingWangAI / SceneMover

Project of Siggraph Asia 2020 paper: Scene Mover: Automatic Move Planning for Scene Arrangement by Deep Reinforcement Learning
MIT License
88 stars 15 forks source link

Question about search algorithm A start #4

Open baifanxxx opened 3 years ago

baifanxxx commented 3 years ago

Hi Wang,

Thank you for your projects and contributions. I follow your work for some time. Recently, I have a question about search algorithm A start which you used. In my simple test, there are two objects, their start point are [32,32],[48,48], two target point are [20,20],[48,48]. When I select the first point to move to target by action A, the path I draw is not the best in real world, may be best with computing cost of the A in discrete space.

https://github.com/baifanxxx/IPM-MovePlanner/blob/main/IPM-MovePlaner/figs/figtest2.png https://github.com/baifanxxx/IPM-MovePlanner/blob/main/IPM-MovePlaner/figs/figtest.png

In my view, to smooth the path point or get the best path in the real, we should adjust the cost function of A or replace A with D. Can you tell me where is the cost function of A in search.cpp?

Best regards, BAI Fan

baifanxxx commented 3 years ago

Hi Wang,

Thank you again for your excellent work. I still hope you help me again. Regarding the A trajectory problem, I think you are using the A algorithm in four directions, up, down, left, and right. In practice, the A algorithm in eight directions may be the best. Excuse me, in search.cpp, how do I modify the direction of each movement of the A algorithm?

Best regards, BAI Fan

HanqingWangAI commented 3 years ago

Hi @baifanxxx. To change the 4-neighbors into 8-neighbors search, you can change the offset array here, and here the number of neighbors.