Go to Polygence Scholars page
Ethan Kou's cover illustration
Polygence Scholar2022
Ethan Kou's profile

Ethan Kou

Henry M Gunn High SchoolClass of 2024Palo Alto, California



  • "State estimation and motion planning for an autonomous competition robot" with mentor Acshi (Dec. 10, 2022)

Project Portfolio

State estimation and motion planning for an autonomous competition robot

Started Apr. 26, 2022

Abstract or project description

Two crucial factors to creating a mobile autonomous robot capable of quickly navigating a flat field are frequent accurate localization and efficient path following. Odometry and computer vision landmarks are two very common methods of localization, but they each have their pros and cons. Odometry has frequent velocity updates at the cost of position error accumulation. Computer vision has very accurate position measurements at the cost of frequency. These two sources of data can be “fused” to create a better position estimate using the Kalman Filter. The Kalman Filter is a full state estimator that can take in localization data of various measurements, frequencies, and certainties to calculate a better estimate of the true position state. Accurate localization allows for path following. Pure Pursuit is a path following algorithm that works by pursuing that furthest point on the path that is within a certain distance from the robot. It controls each state value separately from the others using a Proportional Integral Derivative Controller (PID). Model Predictive Control (MPC) is an algorithm that allows for control of a system. In this case, this system is the robot and its position. When controlling a system to reach the target state, MPC takes all state values into account at once when deciding what to do on each time step. What are the benefits and drawbacks of MPC and Pure Pursuit for use on a mobile robot to quickly reach a target while navigating around obstacles.