Motion-Planning

Table of Contents

• Dynamic Programming (Rock-Paper-Scissors)
• A* and RRT (Maze Runer)
• Infinite Horizon Control & RL
  • on&off TD policy by Q learning (Hill Climbing)
  • PI & VI (Balance Inverted Pendulum)

Rock-Paper-Scissors

Problem

Use Dynamic Programming to produce policy to win against a friend in a game of Rock-Paper-Scissors, knowing that your friend has 50% preference on one of the three options and 25% preference on other two. But you do not know which option that your friend has 50% preference on.

Hyper-parameters

T: Number of matches of Rock-Paper-Scissors in this game. (Default is 100) N_: Density, Number of Total discretized belief states = (N_ * T). Higher N_, more accurate the model is.

Report

Maze Runner

Problem

Use LRTA* and RRT algorithm to let robot move to goal in a 3D virtual environment.

Report

Hill Climbing

Problem

Let a Cart Climb Up the right hill by swing between two up-hill sides.

Report

Balance Inverted Pendulum

Problem

Apply suitable force such that the Pendulum is facing upwards

Report