Axon 06: Localization & State Estimation

Welcome to the Localization & State Estimation Axon. This track covers how an autonomous robot continuously knows its exact millimeter position, velocity, and orientation on the field by fusing wheel encoders, high-speed IMU gyros, and AprilTag computer vision.

Modules in this Axon

1. Wheel Odometry & Gyro Integration

• The Real-World Problem: How does a robot track where it is on the carpet by accumulating wheel rotations and IMU heading angles?
• Key Concepts: Forward kinematics integration, twist accumulation, wheel slip modeling, and systematic odometry drift over time.

2. AprilTag Computer Vision & PnP Pose

• The Real-World Problem: How does a camera turn a 2D image of a planar AprilTag into a full 3D robot coordinate (x, y, z, roll, pitch, yaw) on the field?
• Key Concepts: Pinhole camera model, intrinsic matrix K, Perspective-n-Point (PnP) solvers, camera-to-robot coordinate transforms, and latency timestamp compensation.

3. Extended Kalman Filter (EKF) State Estimation

• The Real-World Problem: How do we combine continuous 100 Hz wheel odometry with intermittent, noisy 30 Hz AprilTag vision measurements into a single rock-solid global pose estimate?
• Key Concepts: State vector x = [x, y, θ]^T, covariance matrix P, process noise Q, measurement noise R, observation model h(x), and standard WPILib SwerveDrivePoseEstimator integration.