Introspective Perception: Learning to Predict Failures in Vision Systems
This addresses the need for long-term autonomous robot operations in complex dynamic environments, though it is an incremental step toward situational awareness.
The paper tackles the problem of enabling robots to self-assess their decision-making reliability in ambiguous situations by proposing a framework for introspective perception that predicts system failures directly from sensor data. The result is demonstrated in vision-based autonomous MAV flight in outdoor environments, showing effective handling of uncertain situations.
As robots aspire for long-term autonomous operations in complex dynamic environments, the ability to reliably take mission-critical decisions in ambiguous situations becomes critical. This motivates the need to build systems that have situational awareness to assess how qualified they are at that moment to make a decision. We call this self-evaluating capability as introspection. In this paper, we take a small step in this direction and propose a generic framework for introspective behavior in perception systems. Our goal is to learn a model to reliably predict failures in a given system, with respect to a task, directly from input sensor data. We present this in the context of vision-based autonomous MAV flight in outdoor natural environments, and show that it effectively handles uncertain situations.