DARPA is exploring a new approach to robotics that embeds intelligence directly into physical materials, reducing reliance on centralized computing, data transmission, and external communication. This effort targets key limitations in current robotic systems, where performance is often constrained by latency, power demands, and connectivity rather than AI capability. DARPA writes:

1. Actuation and Sensing

Currently, every physical task a robot performs, such as grasping, manipulating, and traversing, requires first perceiving its environment (sensing), then analyzing how to interact with the environment (processing), then finally accomplishing the task (actuation). These sensing-processing-actuation loops are among the most environment-dependent components in any robotic system. DARPA seeks information on material-centric approaches for actuation and sensing for unconstrained environments, including but not limited to:

a. Existing state-of-the-art applications of physical intelligence in robots, including their inherent limitations, implementation timescales and projections (Moore’s law equivalent), and driving performance metrics. Of particular interest are those affiliated with soft robots whose degrees of freedom are not limited to fixed joint mechanics.

b. Pathways to develop foundational, high-quality materials, interfaces, and assembly schemes to overcome the above-mentioned limitations and set a precedent for future applications.

c. Novel material-centric architectures with all three modalities—sensing, processing, and actuation—that perceive, differentiate, and send signals based on different state variables (e.g., pressure, ion concentration, temperature, light, etc.).

d. Embedded proprioception at the component or kernel level to provide real-time hardware state information, enabling awareness and autonomy for regeneration, fatigue, or graceful degradation.

e. State-of-the-art technologies for harvesting energy, transducing energy, and storing energy, including limitations, implementation timescale, and performance metrics.

f. Scalable pathways from materials to systems, in order to develop revolutionary technologies to overcome the above-mentioned limitations.

Through a new Request for Information (RFI), DARPA is seeking ideas for materials that can combine sensing, computation, and actuation within the same structure, enabling robots to respond instantly and autonomously in complex or hostile environments. The goal is to create “physical intelligence,” where perception and action are integrated at the hardware level.

The initiative focuses on two main areas: adaptive materials that merge sensing and movement, and embedded computing that allows real-time decision-making within the material itself. DARPA hopes this shift will lead to faster, more efficient, and more resilient robotic systems designed specifically for mission environments rather than human-centered design.