PerceptOR

Applied Perception was part of the DARPA-funded, SAIC-led PerceptOR team, which was selected for all three phases of the PerceptOR program. API led the Active Perception IPT: our responsibilities included software development and integration of all “active” (energy-emitting) sensors in the mobility architecture, including a 2D multi-spectral laser range finder, four 1D laser range finders, and dual 24 GHz radar units. During Phase I of the program, our work focused on conducting sensor trade-off studies and developing individual algorithms for the selected sensors that enabled one or more off-road navigation tasks – functions like obstacle detection, foliage penetrability assessment, and terrain and/or vegetation evaluation.

During Phases II and III, we extended these individual algorithms and developed a framework called a “Density Map” to merge all sensor information into a unified representation of the world. As well, we developed machine learning algorithms for learning traversal cost assessment by observing the behavior of a human driver, and also worked on incorporating domain knowledge about specific objects that are likely to be encountered in the environment – fences, trees, etc.