Lecture 17 - Detection and Pose Estimation

Overview:

In a race, there are usually multiple players. Now that we know how to map and localize in our environment and path plan to a goal point, how do we take into account another moving vehicle? This lecture goes over a vehicle tracking and prediciton framework that can be used to find the pose of and obstacle vheicle and predict the obstacle vehicle’s probable poses in the future.

Topics Covered:

• Single view geometry

• Pose estimation

• Homography

• Trajectory prediction

• Using April tags

Slides:

Links to additional resources:

• Model Based Vehicle Detection and Tracking for Autonomous Urban Driving

• TrafficPredict: Trajectory Prediction for Heterogeneous Traffic-Agents

• The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs

• World Models (paper)

• World Models (interactive)