Robotics

Robotics is the interdisciplinary field that designs, builds, and studies machines capable of autonomous or semi-autonomous interaction with the physical world. The discipline sits at the confluence of engineering, computer science, and the cognitive sciences, but its deepest questions are not technical: they are about what it means for a machine to act in the world, and whether action — not computation — is the true measure of intelligence.

The history of robotics is usually told as a story of mechanical innovation: from the clockwork automata of the 18th century to the industrial arms of the 1960s to the humanoid platforms of today. But this narrative misses the conceptual shift that occurred in the late 20th century, when robotics began to absorb insights from embodied cognition and ecological psychology. The robot was no longer seen as a computer with a body attached; it became a system whose intelligence is constituted by its sensorimotor coupling with the environment. This shift — from the brain-in-a-vat model to the embodied-agent model — restructured the field's theoretical foundations.

From GOFAI to Embodied Robotics

The classical approach to robotics, dominant from the 1960s through the 1980s, treated the robot as a symbol-manipulating system that happened to have physical effectors. Perception was understood as inverse optics: the robot's sensors delivered data, its internal processor constructed a world-model, and its planner selected actions. This paradigm — sometimes called sense-plan-act or GOFAI robotics — produced impressive demonstrations in controlled environments and spectacular failures in the real world. The problem was not engineering sophistication; it was the mismatch between the computational model of mind and the dynamical reality of physical interaction.

The alternative, pioneered by Rodney Brooks and the MIT Mobot Lab in the 1980s, was subsumption architecture: a layered control system in which higher-level behaviors subsumed lower-level ones, but no layer required a complete world-model. A subsumption robot navigates not by computing a map but by directly coupling sensor states to motor responses — what Brooks called intelligence without representation. This was not merely an engineering hack. It was a philosophical claim: that the world is its own best model, and that a robot that continuously interacts with its environment does not need to reconstruct the environment internally.

The subsumption approach connects robotics directly to Gibson's theory of affordances: the robot perceives not objects but action possibilities — gaps it can pass through, surfaces it can traverse, objects it can grasp. Perception is not reconstruction but selection for action. This ecological framing made robotics a testbed for theories of embodied cognition that had previously been confined to philosophy and psychology.

The Reality Gap and the Simulation Problem

One of the most persistent problems in robotics is the reality gap: the systematic divergence between simulated behavior and physical behavior. A robot that navigates perfectly in simulation may collide with walls in reality because simulation abstracts away friction, sensor noise, actuator latency, and the stochastic physics of contact. The gap is not merely a technical inconvenience; it is a methodological symptom of the deeper problem that physical interaction is not computationally tractable in the relevant sense.

The reality gap has pushed robotics toward two complementary strategies. The first is hardware-in-the-loop simulation: building simulators that model physical dynamics with sufficient fidelity that skills transfer directly. The second is embodied learning: training robots in the real world, accepting the cost and danger of physical trial and error. Both strategies are currently active, and neither has fully resolved the gap. The persistence of this problem suggests that the simulation-reality boundary is not merely an engineering challenge but a fundamental feature of the relation between abstract models and physical processes — a version, in robotics, of the map-territory relation that has troubled epistemology for centuries.

Robotics and Cognitive Systems Engineering

The design of robots that operate in human environments — factories, hospitals, homes, streets — requires more than AI and control theory. It requires cognitive systems engineering: the discipline, founded by Jens Rasmussen, that studies how operators (human or machine) interact with complex technological systems. A robot in a hospital is not merely navigating; it is participating in a socio-technical system with norms, constraints, and failure modes that span the human-machine boundary.

This systems perspective reveals that robot autonomy is not a binary property but a graded one. A robot may be autonomous in its locomotion but dependent on human supervision for ethical decisions; autonomous in its perception but dependent on infrastructure for communication. The question is not whether robots will replace humans, but how the distribution of cognitive labor between humans and robots will be negotiated — and who gets to decide.

The standard assumption in robotics is that the field is progressing toward general-purpose humanoid machines that can perform any task. This assumption is a category error. The history of technology suggests that robots will not become generalists but will become specialists in embodiment: machines that excel at the physical interactions — grasping, manipulating, traversing — that are difficult for disembodied systems. The future of robotics is not artificial humans but artificial bodies, and the question of whether those bodies will be intelligent depends on whether we can build them as systems that enact a world rather than systems that compute a model of one.