Normalizing Objectives

[patrick-5546]

Description

• Original PR (make sure diff matches this) Normalizing Each Objective Class's motionCost() local_pathfinding#66
• Resolves local pathfinding issue Explore alternative ways to install OMPL #33

• Added two functions in the parent class: (i) Creates random sample states; (ii) Compares every pair in that sample and picks the maximum motionCost()
• Inherited these functions in each Objective class to find the max_motionCost for each objective function
• Modified the motionCost() function of each Objective class to normalise its cost using max_motionCost

Verification

• Verified using a simple setup example problem
• Haven't included unit tests

Resources

[akash-venkateshwaran]

Hi Patrick, I have completed the normalization. Done the testing of speed obj as well
There an issue in logic for normalization where I got 1.0 for every pair of random samples. I have fixed that bug.
Also there is a problem with SpeedObjectiveMethod.SAILBOT_TIME where no matter how many samples I use I get diff max_motion_cost every time I run it. I don't think its a huge issue. Please check it out

[patrick-5546]

Couple comments

• Local pathfinding tests now take a really long time to run (went from 6s to >1min). Did you observe the same on your machine? Any idea what may be the cause?
• Why do you think SAILBOT_TIME doesn't converge and why isn't it an issue?
• How much latency does setting the num_samples to 2k add for SAILBOT_TIME?
  • It took a lot of time for me to run this individually 10 times (4s vs. 426s)

[FireBoyAJ24]

We will redo normalizing objectives as there are issues with the implementation, with adding pyompl bindings, and the code itself.