MCZhi / Driving-IRL-NGSIM

[T-ITS] Driving Behavior Modeling using Naturalistic Human Driving Data with Inverse Reinforcement Learning
MIT License
192 stars 37 forks source link

what is the relation about the Boltzmann distribution and candidate trajectories? #9

Closed siami233 closed 1 month ago

siami233 commented 2 months ago

Thanks for your good job! In your paper you mentioned,"we assume the distribution is a Boltzmann distribution related to the returns of trajectories." why you use Boltzmann distribution describe the probity of trajectories. In the fig1, When you are evaluating trajectories, I noticed that the horizontal axis of the probability distribution graph on the right contains "lane change left, lane keeping, lane right". What does this mean? Why does "lane keeping" correspond to the highest probability on the vertical axis?

image

image