In order to act intelligently in an open environment, it is necessary to have an ability to learn and predict physical phenomena under different contexts and surroundings. Such dynamic phenomena can be mathematically modeled as stochastic processes.  In this project, we develop novel algorithms and methods for real-time nonparametric learning and prediction of time-varying stochastic processes.  Funded by the Ministry of Science, ICT, and Future Planning (MSIP).

Resources

• CPSLab@SNU
• Publications@Drive

Motion Controller

Publications

"Real-Time Navigation in Crowded Dynamic Environments Using Gaussian Process Motion Control" (14-Choi).

AR-GPMC

14-Choi outlines the creation of the AR-GPMC (auto-regression gaussian process motion controller).

• They get humans to teleop the robot around simulated humans to generate training data.
• The guassian process is on  where  is the number of previous trajectory points and  is given by , polar co-ordinates to an obstacle. Output is the robot control .
• When there are multiple obstacles, it simply weights the resulting controls from each AR-GPMC output.
• If no obstacles, it doesn't move around at all.

Questions

• How do they get it going towards a waypoint? Default algorithm has no control if no human trajectories.
• What is the convex polygon in the internal state pictures/videos?

Guassian Random Paths

Publications

"Gaussian Random Paths - ICRA 2015"

Questions

• At 1:40 why does it seem to be planning through the obstacle?
• How did he get the anchoring points? Past trajectory points?