Vision-Based Reactive Controllers

Abstract

First, this chapter motivates in Section 3.1 cameras as a well-suited choice for implementing reactive instinctive behaviors and recalls basic models and computer vision principles. After that, a novel principle for estimating the Time-to-Collision for single image points is presented in Section 3.2, before Section 3.3 introduces three representations for obstacle avoidance, navigation, and a concurrent execution of both respectively. Finally, Section 3.4 demonstrates how each of the three representations could be utilized for planning.