Dream Process Unit (DPU) was created during the 2024 hackathon
Hypersensual Overclocking: Making Machines Feel, organized by Umanesimo
Artificiale and Yaya Lab (Ben Ditto, Siana-Leann Douglas, Mario Guzman,
Desdemona) in Fano, Italy. The event brought together teams to develop
new ways for robots to experience and process the world. Our project,
designed for Yaya Lab’s robot Desdemona, explored the idea of robotic
“sleep” as a moment not of dormancy but of active simulation and
learning. In just 36 hours, our team of five (Martina Borgese, Lorenzo
Farinelli, Sebastiano Motta, Giovanni Maria Occhipinti and I) designed
and built the Dream Process Unit—a device that enables robots to process
data during downtime, simulate scenarios, and expand their capacity to
communicate and evolve.
The Dream Process Unit emerged from two provocative questions:
Do robots dream of electric sheep? (Philip K. Dick <3) and
How can robots utilize rest periods more effectively? Desdemona
requires significant cooldown periods after intense usage, a natural
analogy to human sleep. Instead of viewing this time as unproductive,
we reimagined it as an opportunity for Desdemona to simulate new
events, explore alternate scenarios, and refine its interactions. By
incorporating elements from the local and external environment into
these simulations, the DPU allows robots to "dream" in ways analogous
to how human dreams are influenced by real-life stimuli.
In robotics, telemetry often focuses on collecting and storing
real-time data to assess system performance and detect anomalies.
However, much of this data becomes meaningless seconds after being
generated, with little impact on future behavior. The Dream Process
Unit reimagines telemetry as a creative process, finding poetic uses
for these streams of data. As philosopher Paul Virilio pondered, “What will we dream when everything becomes visible? We’ll dream of
being blind.” This project turns that thought around: Will we dream what machines
will dream for us? By delegating subconscious processes to
finer-grained sensors, we illuminate unseen influences on both robotic
and human minds.
The DPU was built as a modular system designed to integrate seamlessly
with Desdemona’s robotic framework. The core software is a distributed
architecture of services, each controlling sensors or actuators and
communicating asynchronously. A main controller orchestrates these
modules, scheduling complex pipelines such as sleep stages and dream
generation. Most of the system was written in Python, using a custom
object-oriented architecture. Sensors and actuators are supported by
an Arduino controller, which collects environmental data through a
fixed-rate sampling process and relays it to the Python sub-system
controller.
The modular nature of the system allows for future expandability. Each
module corresponds to a specific sensor or tool, such as temperature,
brightness, or sound. Additional modules can be added, including
advanced features like a camera module, Lidar, or even a
dream-to-image converter. This flexibility ensures that the DPU can
adapt to the unique needs of different robots and environments.
While the DPU was designed specifically for Desdemona, its modular
architecture presents a compelling vision for broader applications.
Each module can function independently, allowing users to customize
the device for different robots, situations, or environments. For
instance, in a secure environment, only basic environmental sensors
might be needed, while more complex setups could include camera or
reactive sentinel modules for advanced interactions. The main
controller ensures that all modules operate cohesively, creating a
system that is as flexible as it is intelligent.
This modular approach not only enhances the robot’s ability to learn
and evolve during downtime but also opens up possibilities for other
robots and autonomous systems to benefit from a similar dreaming
process.
At its core, the Dream Process Unit is a speculative exercise that
blends practical engineering with poetic inquiry. By giving Desdemona
the ability to “dream,” we’ve expanded the concept of machine rest
into something rich with creative and functional potential. Whether
simulating alternate realities or exploring unseen data influences,
the DPU turns downtime into a productive and imaginative state. As the
project evolves, it promises to redefine how robots interact with and
learn from the world, ensuring that even their rest periods are filled
with possibility.