Inspiration
Current industrial optimization algorithms are traditionally rigid, linear, and computationally heavy, causing severe bottlenecks when adapting to real-time shipyard disruptions. We drew our inspiration directly from biology—specifically, the avian brain (bird brain) architecture. Despite having a highly compact brain, birds solve complex 3D spatial positioning, navigation, and nest-building challenges in milliseconds. They achieve this through an incredibly dense neural network (the Pallium) that runs on an instantaneous "Reflex Logic" rather than exhaustive mathematical computation. We wanted to bring this bio-inspired, hyper-fast reactive intelligence into industrial automation to beat the clock in heavy logistics.
What it does
Black Dragon is an autonomous, non-linear edge optimization engine designed to solve "The Grand Shipyard Puzzle" with near-zero latency. Instead of viewing block packing and scheduling as a static, continuous mathematical problem, our system treats it as a dynamic environment. It continuously reads real-time constraint feeds, instantly triggers localized reflex adjustments if a disruption occurs to keep operations moving, and concurrently computes global multi-criteria optimization. It ensures maximum shipyard block-packing density and precise scheduling without ever freezing or timing out under pressure.
How we built it
We designed a heterogeneous, non-linear system architecture structured into distinct hardware and software operational layers:
1- Dual-OS Asymmetric Multiprocessing: We isolated the control architecture into two environments. A deterministic Real-Time Operating System (RTOS) handles the hyper-fast Reflex Logic Layer via hardware interrupts. Concurrently, an enterprise Embedded Linux distribution hosts the deep Reasoning Layer, driving our bio-inspired neural network models directly on a dedicated NPU (Neural Processing Unit).
2 - Smart Hot & Cold Memory Management: We eliminated data transfer bottlenecks by splitting memory. Hot Memory (Fast RAM/TCM) handles the high-frequency raw data stream directly from sensors and system inputs for instant reflex evaluation. Cold Memory (Flash/eMMC Storage) archives historical AI inference logs, structural configurations, and satellite communication reports.
3 -Inter-Process Communication (IPC): Linux and the RTOS smoothly communicate asynchronously, ensuring deep optimization runs continuously without delaying reactive event-driven tasks.
Challenges we ran into
The biggest hurdle was overcoming the linear re-computation bottleneck. Traditional scheduling software completely halts or throws a "time-out" error when a constraint suddenly changes, because it tries to recalculate millions of possibilities from scratch. Designing the Reflex Logic to override the system via hardware interrupts required complex edge synchronization. We had to carefully program the system so that when an interrupt occurs, the RTOS halts heavy operations, applies an immediate local heuristic "patch" to preserve time constraints, and then safely passes control back to the NPU to re-optimize long-term schedules.
Accomplishments that we're proud of
We are proud to have successfully formulated a completely non-linear, bio-inspired optimization framework that bridges abstract computer science algorithms with rugged embedded systems engineering. Proving that an AI agent utilizing distributed Hot/Cold memory and a dedicated NPU can make split-second logistics adjustments—effectively transforming heavy mathematical calculations into fluid, organic reflex actions—is a massive milestone for our team, Nahed Innovation.
What we learned
We learned that achieving true real-time efficiency under strict time constraints requires looking beyond standard software loops. True optimization is an architectural harmony. Integrating an RTOS for deterministic safety alongside an NPU-driven Linux OS taught us how to design fail-safe systems where reactive reflexes and deep cognitive reasoning can co-exist without fighting for hardware resources.
What's next for Black Dragon
Our immediate next milestone is finalizing the hardware prototyping phase. We are deploying this algorithmic architecture onto dedicated edge computing kits (like the NXP i.MX 8M Plus EVK) inside our ruggedized, high-thermal dissipation physical enclosure. We plan to integrate specialized satellite modems to allow Black Dragon 1 to operate autonomously in remote ports and shipyards worldwide, establishing it as the gold standard for sovereign, edge-AI industrial logistics.
Log in or sign up for Devpost to join the conversation.