Paul Vernaza

About

Paul Vernaza is a roboticist whose research lies at the intersection of motion planning, machine learning, and perception for autonomous systems operating in complex, unstructured environments. His most influential work tackles the challenge of legged locomotion over rough terrain, where he pioneered a search-based planning approach (53 citations) that generates complete joint trajectories for quadrupedal robots. Vernaza also made significant contributions to autonomous terrain classification, developing an online, self-supervised method using discriminatively trained submodular Markov random fields (51 citations) that enables robots to segment images into obstacle and ground patches without extensive manual labeling. His work extends to cooperative localization, where he demonstrated how teams of mobile robots can estimate relative poses using only audible acoustic sensing (20 citations). Vernaza further advanced planning under topological constraints through his beam-graph framework (14 citations), which modifies standard graph-based navigation to incorporate topological information. His research portfolio also includes robust GPS/INS-aided localization and mapping, as well as efficient dynamic programming for high-dimensional motion planning using spectral learning of value function symmetries. A former member of the University of Pennsylvania's RoboCup legged soccer team, Vernaza's work has been foundational in bridging perception, planning, and control for autonomous robots operating in challenging real-world environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Search-based planning for a legged robot over rough terrain

   53 citations · 2009
2. 2
   Online, self-supervised terrain classification via discriminatively trained submodular Markov random fields

   51 citations · 2008
3. 3
   Cooperative relative robot localization with audible acoustic sensing

   20 citations · 2005
4. 4
   Planning under topological constraints using beam-graphs

   14 citations · 2013
5. 5
   Robust GPS/INS-Aided Localization and Mapping Via GPS Bias Estimation

   8 citations · 2008
6. 6
   Efficient dynamic programming for high-dimensional, optimal motion planning by spectral learning of approximate value function symmetries

   5 citations · 2011
7. 7
   Learning and planning high-dimensional physical trajectories via structured Lagrangians

   3 citations · 2010
8. 8
   The University of Pennsylvania Robocup 2004 legged soccer team

   2 citations · 1987

Key Collaborators

Contact & Links