Thursday, 2023/03/30

  • CS/AI
  • C
  • TITECH
  • Switch Language
    • ja日本語 (Japanese)
    • enEnglish

Shimosaka Research Group

pursuing MIUBIQ (machine intelligence in UbiComp Research)

  • Home
    • Members
    • Location
  • News
  • Projects
  • Publications
  • Awards
  • Archives
    • Codes
    • Datasets
Navigation
Projects Driving behavior modeling at un-signalized intersection with inverse reinforcement learning on sequential MDPs

Driving behavior modeling at un-signalized intersection with inverse reinforcement learning on sequential MDPs

2021/07/12 | Projects | 1481 views |
Driving behavior modeling (DBM) is widely used in the intelligent vehicle field to prevent accidents, which predicts actions that vehicles should take to optimize safe driving behaviors. According to some statistics, accidents easily happen at un-signalized intersections, modeling driving behavior at such places is of great importance.

Inverse reinforcement learning is gaining popularity in recent DBM tasks, which learns the reward functions of reinforcement learning from real driving data. However, current inverse reinforcement learning-based DBM methods fail to predict proper behaviors at the un-signalized intersections in the aspects of smoothness and stopping behavior by just using a single Markov decision process (MDP). When modeling driving behavior at the intersection, there are two problems with the single-MDP method. If planning behavior by maximizing the reward, the car tends to stop around the stop line for a long time to earn as much reward as possible. While by stochastic policy from reward expectations, stopping time around the stop line becomes shorter but unusual acceleration will happen. In other words, it is difficult for current methods to maintain a balance between the stability of the speed and the rationality of the stopping behavior at the stop line. That is abnormal for real human driving behavior.

In this project, for smoothening the driving path, we divide the target intersection into three parts according to the position of the stop line. A sequential MDPs structure is proposed to model the driving behavior at the intersection separately. And a Poisson model is used to simulate the stopping behavior around the stop line close to human driving behavior. The experimental results show that our model generates a smoothed driving path with reasonable stopping behavior at the un-signalized intersection.

Publications

Shaoyu Yang, Hiroshi Yoshitake, Motoki Shino, Masamichi Shimosaka.
Smooth and Stopping Interval Aware Driving Behavior Prediction at Un-Signalized Intersection with Inverse Reinforcement Learning on Sequential MDPs.
The 2021 IEEE Intelligent Vehicles Symposium (IV21)

  • tweet

Comments are disabled for this post

Social Networks

  • twitter
  • rss

Recent News

  • Device-Free Multi-Person Indoor Localization Using the Change of ToF 2023/03/03
  • Presenting our paper on Device-Free Multi-Person Indoor Localization Using the Change of ToF at PerCom2023 2023/02/28
  • Our paper on Efficient Adaptive Beacon Deployment Optimization for Indoor Crowd Monitoring has been published in IMWUT. 2023/01/24
  • Presenting our paper on Robust Continuous MaxEnt IRL with RRT at IV2022 2022/06/09
  • Presenting our paper on Efficient Indoor Localization Model Construction by Sequential Recommendation of Data Gathering Position based on Bayesian Optimization at IPIN2021 2021/11/29
  • Adaptive incremental beacon placement optimization for crowd density monitoring applications 2021/11/01
  • Presenting 2 papers at ACM SIGSPATIAL 2021 2021/11/01
  • Fine-grained Urban Dynamics Prediction using Large-Scale Mobile Phone Location Data 2021/10/05
  • Robustifying Wi-Fi localization by Between-Location data augmentation 2021/09/28
  • Our paper on robustifying Wi-Fi localization by “Between-Location” data augmentation has been published in IEEE Sensors Journal 2021/09/17

Search

Copyright 2015 · Shimosaka Research Group at TITECH