Assessment of Navigation Using a Hybrid Cognitive/Metric World Model

Abstract : One goal of the US Army Research Laboratory's Robotic Collaborative Technology Alliance is to develop a cognitive architecture that would allow a robot to operate on both the semantic and metric levels . As such, both symbolic and metric information would be interpreted within the robot's world-model with subsystems providing or interpreting the appropriate type of information. In December 2013, an integrated research assessment was conducted to evaluate progress toward this goal. The robot's ability to receive instructions in structured text, and to interpret those instructions in the context , in both camera and laser detection and ranging images, was applied to the goal of navigating around a building. In this report, the robot's performance is evaluated with 3 measures: 1) grading of the robot's performance during real-time operation, 2) evaluation of planning given the perceptual data within the world model after the testing , and 3) after- the- fact diagnosis of the cause of failed runs by developers of the technology.