Localization and Mapping
Mutual Information for Exploration
Depth Sensing and Perception
A broad range of next-generation applications will be enabled by low-energy, miniature mobile robotics including insect-size flapping wing robots that can help with search and rescue, chip-size satellites that can explore nearby stars, and blimps that can stay in the air for years to provide communication services in remote locations. While the low-energy, miniature actuation, and sensing systems have already been developed in many of these cases, the processors currently used to run the algorithms for autonomous navigation are still energy-hungry. Our research addresses this challenge as well as brings together the robotics and hardware design communities.
We enable efficient computing on various key modules of other autonomous navigation systems including perception, localization, exploration and planning. We also consider the overall system by considering the energy cost of computing in conjunction with actuation and sensing.
Many motion planning and control algorithms aim to design trajectories and controllers that minimize actuation energy. However, in low-energy robotics, computing such trajectories and controls themselves may consume a large amount of energy. We develop algorithms that optimize this trade-off.
Computing mutual information between the map and future measurements is critical to efficient exploration. Unfortunately, mutual information computation is computationally very challenging. We develop new algorithms and hardware for efficient computation of mutual information, and demonstrate real-time computation for the whole map in a reasonably-sized map.
Depth sensing is a critical function for robotic tasks such as localization, mapping and obstacle detection. State-of-the-art single-view depth estimation algorithms are based on fairly complex deep neural networks that are too slow for real-time inference on an embedded platform, for instance, mounted on a micro aerial vehicle. We address the problem of fast depth estimation on embedded systems.
Autonomous navigation of miniaturized robots (e.g., nano/pico aerial vehicles) is currently a grand challenge for robotics research, due to the need for processing a large amount of sensor data (e.g., camera frames) with limited on-board computational resources. We focus on the design of a visual-inertial odometry (VIO) system in which the robot estimates its ego-motion (and a landmark-based map) from on-board camera and IMU data.