DARPA wants robots that compute with their bodies' material
The US government wants to develop a new material with processing built in for robots operating in extreme defence environments.
The Defense Advanced Research Projects Agency (DARPA) said the project, labelled 'Rethinking robotics with physical intelligence', will research physical materials that can sense and process information to improve robotics systems.
Defence hardware differs from commercial robotics as it must excel in "extreme, unpredictable, and adversarial environments."
"Performance is not defined by how much data a system can process, but by how quickly and reliably it can respond," the agency said, requiring a shift in the system design.
In a request for information (RFI) notice published on April 29, the agency outlined plans for “a new class of materials capable of intermixed sensing, adapting, and acting in real time without relying on continuous external computation or communication links.”
Collapsing the loop
DARPA, which develops new technologies for the US military, called for external researchers to approach it with ideas for the programme, aiming to “fundamentally reshape what robotic systems can do.”
Programme Manager Julian McMorrow cited the need to close processing ‘loops’ and reduce power demands if robotic defence systems are to operate in extreme environments with “limited communication and little opportunity for human intervention.”
He said: “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.” Embedding compute into the materials themselves is the goal for closing the loop.
DARPA's RFI process, logged as DARPA-SN-26-76, will culminate with an invite-only workshop this summer. The response deadline is May 27, 2026.
Smart materials
Interest in smart materials and “soft robotics” is far from new. University of Cambridge researchers with a similar goal to DARPA combined ionic electro-active polymers and shape memory polymers in 2015, but AI advancements have led to fresh interest in the topic.
In February 2026, the University of Birmingham published research on “speed-sensitive metamaterials”, inspired by the physical properties of rice, that could improve a robot’s abilities in harsh environments and for delicate tasks.
See also: The Big Interview: Johanna Hutchinson, CDO, BAE Systems
In 2024, an EU funded SMART project also finished a four-year research programme on “self-healing soft robots”, using AI and machine learning tools to develop materials integrated with sensing and actuation tech to detect and repair their own damage.
DARPA said its research would focus on materials capable of both sensing and actuation, and integrated with “elements of control”, explaining that “instead of routing information through centralized processors, future systems could respond through their physical design.”
It added that the robots do not need to be humanoid and welcome "designs that are smaller, larger, softer, or structurally unconventional, prioritizing performance and adaptability over familiarity."
Robotics and AI
AI advancements have also led to billions being poured into robotics startups by VC firms and legacy tech companies alike.
In 2026, some of the largest robotics startup funding rounds include Skild AI’s $1.4 billion Series C, Mind Robotics’ $500 million Series A, and Apptronik’s $520 million Series A extension.
Chip giant NVIDIA has also highlighted robotics and edge AI as an area of interest, operating its own Robotics Research arm and working with Disney and Google on a virtual robotics engine. In March 2026, CEO Jensen Huang said physical AI progress would see “every industrial company” become a robotics company.
NVIDIA is building smaller and more powerful edge AI chips, like its Jetson Nano hardware, to let machines run autonomously without needing to connect to the internet or another machine, but projects like DARPA’s could effectively remove the need for such chips.