Dr. Nisar Ahmed not only considers the technological challenges of autonomous robotics, but also, the philosophical. Contrary to the popular sentiment that advancements in autonomous robots will replace human activities, Ahmed believes it will merely support and enhance them.
Ahmed’s involvement in robotics grew from an interest in control systems, particularly as they pertain to the human body. Originally focused on biomedical engineering, Ahmed’s graduate work at Cornell University introduced him to robotics and its controls.
To Ahmed, autonomous vehicle research was alluring because it posed “really enjoyable challenges.” After receiving his Ph.D. in Mechanical Engineering, Ahmed was attracted to RECUV because of its extensive experience with outdoor field operations and its ability to “do everything from circuit boards to mission design.”
While much of RECUV’s research concerns improving the technological performance and capabilities of autonomous vehicles, Ahmed’s research centers on promoting “partnerships” between humans and robots:
“People are so focused on the nitty-gritty technical things of how to make an air vehicle autonomous, they miss the bigger point that drones are still serving people. You can never fully remove the human out-of-the-loop; you can just add more loops beneath the human. When a drone doesn’t do something correctly, what’s the reason? User input? Incorrect software? How do you have a robotic ‘teammate,’ almost like your partner? These are questions that aerospace is running into very fast.”
Ahmed's research team using infrared tracking data to model an object's movement in a 3D simulation environment in real time.
To answer these questions, Ahmed has directed his research at developing estimation algorithms that promote “cooperative intelligence” between man and machine. Cooperative intelligence is the integration of human knowledge into autonomous robotic decision-making to compensate for intrinsic limits in robotic sensing, actuation and processing.
Cooperative intelligence could improve the effectiveness of teams of humans and robots engaged in search-and-rescue, surveillance and exploration. For example, an autonomous robot conducting search-and-rescue could take suggestions from human rescuers on-site (such as, “try looking for survivors under that wall behind you!”) to refine and expedite its search. In this way, humans are not just providing commands to the robots; they are acting like “sensors” that provide information that is incorporated into the robots decision-making process.
Human knowledge can be helpful, but it is not infallible. How do robots take human suggestions into account, while weighing the possibility that those human inputs are incorrect? How do robots objectively process human beliefs and intuitions?
Nisar's research team running human-robot interaction experiments, where the human can act as a sensor, providing statements like, “I think Zhora (a robot target) is behind the desk,” (shown above). The ‘codebook’ of statements above is a first step towards natural human-robot interfaces.
Translating human directives into information that robots can understand is a challenge in of itself. To Ahmed, it is imperative that a human operator not be required to have a Ph.D. in robotics in order to collaborate with a robot. Rather, the robot must be able to account for a diversity of human users with different backgrounds (military, civilian, etc.) and communication styles. Ahmed and his research team are currently developing semantic models that they hope will account for this variety.
Aside from the technical challenges, Ahmed foresees societal obstacles to the implementation of autonomous technology:
“Something in human psychology is inherently distrustful of autonomy at first (consider Google cars). The challenge then is: how do you design a robot to be trusted? What the public needs to understand is that there are legitimate safety issues. However, people shouldn’t be afraid about losing their jobs tomorrow. The more they understand about the limitations of autonomy, the more they will realize that robots will support the role of humans, not replace it.”
Ahmed continues by explaining the opportunities for human productivity and intellectual growth that are enhanced by autonomous robots:
Challenges with human-robot communication can make robots hard to tame.
“In the long-run, autonomous robots offer a justification for improving our education system and promoting the higher-level thinking that humans are naturally good at. Instead of replacing 1000 workers with 100 robots, we should focus on finding a way for those 1000 workers to make better use of those 100 robots. It’s about cooperation, not competition.”
In collaboration with several RECUV graduate and undergraduate students, Ahmed is writing a workshop paper on semantic modeling to be presented at the 2015 Robotics: Science and Systems Conference. By August, Ahmed hopes to have a fully functioning hardware demonstration of the cooperative-intelligence robotic test bed “which will be used to support work that will be submitted to several other leading robotics conferences and journals.”
-Written By: Ari Sandberg, Intern