Active Classification of Moving Targets with Learned Control Policies
This addresses the challenge of integrating black-box classifiers into active information-gathering for robotics, enabling more efficient and scalable classification of moving targets in dynamic environments.
The paper tackles the problem of a drone actively collecting semantic information to classify multiple moving targets by computing control inputs for informative viewpoints, using a black-box classifier like a deep neural network. The result is a novel attention-based architecture trained via Reinforcement Learning that outperforms baselines, generalizes to unseen scenarios, scales to many targets, and handles different movement dynamics.
In this paper, we consider the problem where a drone has to collect semantic information to classify multiple moving targets. In particular, we address the challenge of computing control inputs that move the drone to informative viewpoints, position and orientation, when the information is extracted using a "black-box" classifier, e.g., a deep learning neural network. These algorithms typically lack of analytical relationships between the viewpoints and their associated outputs, preventing their use in information-gathering schemes. To fill this gap, we propose a novel attention-based architecture, trained via Reinforcement Learning (RL), that outputs the next viewpoint for the drone favoring the acquisition of evidence from as many unclassified targets as possible while reasoning about their movement, orientation, and occlusions. Then, we use a low-level MPC controller to move the drone to the desired viewpoint taking into account its actual dynamics. We show that our approach not only outperforms a variety of baselines but also generalizes to scenarios unseen during training. Additionally, we show that the network scales to large numbers of targets and generalizes well to different movement dynamics of the targets.