Papers
56
Total Citations
1,842
H-Index
22
About
Matthew Johnson-Roberson is a pioneering roboticist and computer scientist whose work spans underwater autonomous systems, simultaneous localization and mapping (SLAM), and autonomous driving. His early and highly influential research established robust frameworks for large-scale 3D reconstruction from underwater robotic surveys, earning 256 citations and laying the groundwork for marine robotics applications ranging from benthic habitat analysis to archaeological discovery. His underwater mapping work has been transformative: collaborating on high-resolution documentation of submerged archaeological sites, he demonstrated that autonomous underwater vehicles (AUVs) could produce geometrically accurate photomosaics where traditional methods fail, with related papers collectively accumulating nearly 400 citations. Johnson-Roberson has also made substantial contributions to autonomous driving and machine learning. His "Driving in the Matrix" work cleverly leveraged synthetic virtual environments to replace costly human-annotated training data, addressing a fundamental bottleneck in deep learning pipelines. More recently, his BiTraP pedestrian trajectory prediction framework (184 citations) advanced safety-critical navigation by enabling robots to anticipate human movement with multi-modal goal estimation. Additional contributions include robotic grasping under partial observation and sonar-based SLAM for ship hull inspection. Across domains, his research reflects a consistent commitment to deploying intelligent robots in challenging, real-world unstructured environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5Mind the gap - robotic grasping under incomplete observation109 citations · 2011
- 6Mapping Submerged Archaeological Sites using Stereo-Vision Photogrammetry107 citations · 2013
- 7Pose-Graph SLAM Using Forward-Looking Sonar98 citations · 2018
- 8
- 9Attention-based active 3D point cloud segmentation48 citations · 2010
- 10