It is hard to be a robot in the wild
As discussed in Science Robotics, the family movie The Wild Robot illustrates four rigors of real-world field robotics.
The Wild Robot is one of the best films of recent memory- it's an animated film version of the classic children’s book. A bipedal humanoid robot named Roz (a call back to Rossum’s Universal Robots and the origin of the word “robot”) is stranded on an uninhabited island. Her only companions are the local wildlife, including a gosling who has imprints on her as his mother (big shout out to Konrad Lorenz and his influence on behavioral robotics). Roz is intended to be an agricultural robot working in very structured environments, but humorously adapts to the unpredictable demands of the wild. As discussed in Science Robotics, which you can read here behind the paywall, the movie captures many of the challenges that real- world robots encounter when operating in the wild. See more of my Science Robotics science fiction-science fact column here.
Four big challenges of working in the wild versus a laboratory that are highlighted in the movie are:
the lack of structure or predictability in the robot’s working envelope (area it works in).
reliability over long durations of time (weeks, months or even years- think Mars rovers- of being in service)
giving verbal instructions via generative AI is harder because there are unexpected corner cases and there may not be a repository of procedural data that can serve as a how-to template
having a robot be authentic rather than rely on social engineering to manipulate others
The question of how can a robot be continuously deployed in environments that are unstructured and dynamically changing? is a hard one. The movie answers it with Roz endearingly saying "Sometimes, to survive, we must become more than we were programmed to be.” Of course we roboticists do not know how to do this, though it does seem related to work in emergent behaviors and resilience- both big topics.
For more discussions of robots and the wild in Science Robotics see:
My book Disaster Robotics goes into detail about field conditions, how to categorize them and how to test and evaluate. One of our recent papers "Water, lava, and wind: Lessons learned for field robotics and human factors research during real world disasters" S. Camille Peres, Ranjana K. Mehta and Robin R. Murphy describes the HRI aspects.