Sensor fusion using a selective sensor framework to achieve decision and task execution

In this paper, the role of multi-sensor data fusion to achieve better interaction levels for Human Robot Collaboration is highlighted. The decision framework uses a sensor-map that is akin to our brain-map and takes into account the pre-processed information streamed by the sensors and groups the sensors and sensor data before applying to the fusion algorithm to derive a confidence level. The task-map is a dynamic structure that holds relevant details between the tasks and the corresponding sensors involved to execute the task. The confidence generated is again compared against real-world constraints before performing the interaction. In the event of uncertainty, the robot can query the user for assistance before executing any further action, thus demonstrating key interaction with the human.