Daniel Maturana
Carnegie Mellon University, Pontificia Universidad Católica de Chile
Papers
8
Total Citations
4,522
H-Index
6
About
Daniel Maturana is a leading researcher at the intersection of 3D perception, autonomous navigation, and deep learning for robotics. His seminal work, **VoxNet** (2015, over 3,500 citations), introduced a groundbreaking 3D Convolutional Neural Network that enabled real-time object recognition directly from LiDAR and RGB-D point clouds, fundamentally advancing how robots perceive their environment. Building on this, Maturana made critical contributions to autonomous flight in GPS-denied environments, developing robust visual odometry and mapping systems using RGB-D cameras (2012, 2016) that allow drones to navigate and map indoor spaces without external positioning. His research also extends to reinforcement learning, where he improved continuous control policies by modeling action bounds with the Beta distribution (2017), and to single-image 3D scene layout estimation for real-time obstacle detection. More recently, he has explored learning-based ego-motion estimation with novel homomorphism-based losses and drift correction, as well as deep inverse reinforcement learning for predicting off-road vehicle trajectories. Maturana’s work has been instrumental in making 3D deep learning practical for real-world robotics, bridging the gap between perception and autonomous action.
Research Focus
Key Achievements
Top Papers
- 1VoxNet: A 3D Convolutional Neural Network for real-time object recognition3,579 citations · 2015
- 2Visual Odometry and Mapping for Autonomous Flight Using an RGB-D Camera610 citations · 2016
- 3
- 4
- 5
- 6
- 7
- 8Indoor Mobile Robotics at Grima, PUC2 citations · 2011