Desai’s and Yanco’s paper discusses a sliding scale autonomy system that allows autonomy levels to be created and changed on the fly. This paper specifically addresses the how the level of human robot interaction varies, measured by the amount of intervention required. In situations where the robot requires a combination of available autonomy modes, a sliding scale autonomy can be used to provide intermediate autonomy levels on the fly, thus providing a great deal of flexibility and hence allowing optimum usage of the system.
The robot’s behavior can be determined using any robot architecture, ranging from reactive control to a hybrid architecture. The extensibility of this model is nice in that it can be used in different architectures. This also addresses the concern presented in Parasuraman’s “A Model for Types and Levels of Human Interaction with Automation” where the human operator works with the autonomous system in tandem.
The force field model provides a mechanism for the operator to adjust the speed of the robot depending on how effective the operator is. Modeling the different variables that are included in the operation of a robot creates a nice visualization of complex data. This approach allows the transition from a human controlled robot to a fully autonomous robot. This is a great method for simulating a trust relationship with the robot because the gradual slide allows the operator to gradually increase the autonomy based upon how the operator is using the robot.
The other interesting piece of the paper is in regards to how the robot avoided obstacles. It seems that the robot looks for the open space and therefore any robot behavior generating rotate and translate would return similar results. In looking at the architecture, the operator gives commands to the system. The robot also chooses its own behaviors based on the data. Both sets of information are sent to a behavior arbitrator which then decides the appropriate behavior to act on. This is a similar model is in Goodrich’s “Experiments in Adjustable Autonomy”, where there is an interface agent that sits between the human operator and the autonomous system. This approach allows both the human operator and the robot to input their behaviors and delegate the choice to an independent entity. This is also part of the interaction allowing for the human operator to trust that the robot is making good decisions.
Reference:
- Desai, Munjal & Yanco, Holly A. “Blending human and robot inputs for sliding scale autonomy.” <http://www.researchgate.net/publication/4177613_Blending_human_and_robot_inputs_for_sliding_scale_autonomy>
