Researchers, materials scientists, and engineers have eight days to help shape what could become the U.S. military's next major robotics program. The Defense Advanced Research Projects Agency's Microsystems Technology Office issued a Request for Information on April 27 under Special Notice DARPA-SN-26-76, titled "Materials for Physical Compute in Untethered Robotics," seeking concepts for robots whose sensing, computation, and actuation would be embedded in their physical materials rather than routed through central processors. Responses are due May 27, 2026, at 2 p.m. Eastern Time.
The submission window matters because DARPA will use responses to select participants for an invite-only in-person workshop in Arlington, Virginia, tentatively scheduled for June or July 2026. That workshop is expected to inform a formal Broad Agency Announcement — the mechanism through which DARPA distributes research contracts. Researchers whose concepts align with the agency's priorities will be invited to present; those who do not submit by May 27 are excluded from consideration.
The Bottleneck DARPA Is Trying to Remove
Every robot fielded today, from industrial arms on factory floors to wheeled reconnaissance platforms, shares the same basic architecture: sensors collect data, transmit it to a central processor, the processor computes a response, and actuators execute the result. That loop is fast by human standards, but not instantaneous. The gap — ranging from microseconds to milliseconds depending on the system — represents latency, power consumption, and a point of vulnerability.
"Today's robots are often limited by the need to sense, process, and act as separate steps," said DARPA Program Manager Julian McMorrow in the agency's announcement. "We are interested in collapsing that loop by embedding intelligence directly into the hardware, so systems can respond in real time without relying on constant data movement."
In contested military environments where GPS can be jammed and radio-frequency communications disrupted, the centralized architecture creates a different kind of problem: a robot that depends on continuous datalinks to function can be disabled without touching the machine itself. The RFI document is explicit on this point, noting that current robotics approaches — increasing sensor density and data transfer volume — come at the cost of reduced energy efficiency, speed, and information security.
What DARPA Means by Physical Intelligence
The RFI defines physical intelligence as the encoding of sensing, actuation, control, learning, adaptation, and decision-making directly into a robot's body — not in software running on a processor, but in the geometry, compliance, and material properties of the structure itself.
The approach has an established name in academic robotics: morphological computation. The concept, developed steadily since the early 2010s, holds that a body's physical structure can itself perform computation, offloading cognitive work from a central processor. Researchers including Helmut Hauser, Professor of Soft Robotics and Morphological Computation at the University of Bristol, and Josie Hughes, head of the Computational Robot Design and Fabrication Lab at EPFL, have published foundational work on how material properties and structural geometry can replace or supplement centralized control.
Nature provides the proof of concept. A sea anemone contracts when touched not through nervous deliberation but through the mechanics of its tissue. Research has shown that the human hand performs a significant part of its grip computation through the passive dynamics of tendons and skin before the brain is consulted.
DARPA's RFI asks what it would take to engineer those properties deliberately and at defense-relevant scale. The candidate materials are not exotic in principle, though difficult in execution: shape-memory polymers that shift geometry in response to temperature or stress, piezoelectric sensors that convert mechanical pressure directly into electrical signals, and pneumatic actuators capable of running local feedback loops without central oversight.
Two Tracks, One Goal
The RFI identifies two specific research areas. The first centers on actuation and sensing — developing materials and structures that combine sensing, processing, and actuation in a single physical substrate, enabling robots to perceive and interact with their environment faster and more efficiently than current designs allow.
The second targets what DARPA calls dynamic and adaptive closed-loop compute: embedding computational capability directly within a material so that real-time decisions can occur with minimal latency, reduced power draw, and continuous adaptation to changing conditions. DARPA is particularly interested in systems that can process multiple environmental inputs simultaneously — pressure, temperature, and light among them — without routing that information through conventional processors.
The agency has stated it is not looking for incremental improvements to existing architectures, software-based approaches, or solutions built around conventional central processing units or graphics processing units. Concepts that merely describe existing capabilities without a vision for advancing the field will not be reviewed.
Why the Military Needs This Now
The operational context is near-peer conflict, defined by adversaries capable of jamming GPS signals, disrupting radio communications, and degrading the datalinks that current autonomous systems depend on. A drone that requires continuous communication to function is, in that environment, a single radio-frequency jammer away from uselessness.
A robotic system whose responses are encoded in its materials — and whose sensing and actuation loops are closed locally — is substantially harder to disable remotely. It is also lighter and more power-efficient, because every gram and watt dedicated to a central processor can be redirected to the mission. At the smallest scales — insect-sized or smaller — conventional microprocessors become impractical regardless of communications availability. Physical intelligence may be the only viable architecture for truly miniaturized autonomous systems.
A Long Road From Workshop to Field
The RFI is explicitly exploratory. DARPA is not distributing research contracts at this stage; it is mapping the landscape, identifying which researchers are working on what problems, and scoping where the technical barriers lie. The solicitation asks respondents to address fabrication scalability, durability under field conditions, and the theoretical limits of computation embedded in matter.
The path from the summer workshop to a deployable robotic system is measured in years, not months. Materials science, fabrication technology, and the theoretical frameworks for programming physical structures would all need to advance in parallel. Researchers in the field have been candid that the distance between laboratory proof-of-concept demonstrations and a machine robust enough for uncontrolled field use remains significant.
DARPA's involvement is itself a signal worth noting. The agency's interest has historically preceded transformative technologies by years or decades; the internet, GPS, and speech recognition each passed through DARPA programs before reaching widespread use. Whether physical intelligence follows that trajectory depends on what the research community submits by May 27 — and whether the materials science turns out to be as programmable as the biology suggests it could be.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
