Apex-Gen Race & Flow Simulator

Inspiration

Our inspiration stems from the high-stakes, real-time competition found in advanced mobility sports like Formula E and high-speed autonomous drone delivery races. Current simulation tools often treat vehicle dynamics and strategic decision-making separately, leaving a gap. We were inspired to design a lightweight simulator that could organically blend micro-level physics (tire degradation, energy consumption, drag) with macro-level strategy (when to pit, overtake, or conserve power), all while enforcing complex regulatory rules. We wanted to build a digital competition where the winner is decided not just by speed, but by compliant, intelligent strategy.

What it does

The Apex-Gen Simulator is designed to model the competitive interaction of a high volume of moving agents in a constrained environment. It functions as a single, unified platform that:

Simulates Physics: Accurately models vehicle energy states, velocity, and trajectory based on defined road or air geometry.

Runs Strategic AI: Each agent utilizes an integrated strategy layer to make real-time, non-deterministic decisions (e.g., initiating an overtake or adjusting its power curve) based on proximity to rivals and resource levels.

Enforces Compliance: Continuously monitors track limits, regulated energy usage, and fair play, applying real-time penalties.

Outputs a Live, Compliant Leaderboard: Provides a dynamic, legally verified output, determining the final ranking based on time and regulatory infringements. This moves the simulation result from a raw time trial to a genuine competitive result.

How we built it (Conceptually)

The Apex-Gen concept is designed around a highly optimized, modular stack. The core of the design relies on a parallelized architecture where the Physics Kernel (handling complex differential equations for movement) is separated from the Strategic AI Layer (handling high-level decision logic). We propose using Python for the flexible AI modeling and a high-performance language like C++ for the core Physics Kernel to ensure rapid, real-time calculation.

The regulatory function is handled by a separate Compliance Engine which utilizes highly efficient spatial data structures (like a Quadtree) to instantly query the positions of all agents and detect infractions without bottlenecking the main simulation loop. This strict separation of concerns is fundamental to the simulator's theoretical performance.

Challenges we ran into (Conceptual)

The main theoretical challenges in designing Apex-Gen involve managing complexity and maintaining performance at scale:

Parallel State Synchronization: Ensuring that the high-speed Physics Kernel and the high-level Strategic AI Layer remain perfectly synchronized without causing bottlenecks or race conditions is a major design hurdle.

Realistic Non-Determinism: Designing the Strategic AI to be "competitive" and not merely follow fixed, scripted routes. Introducing realistic elements like tire degradation and stochastic power consumption $\left( E_{loss} \propto v^2 + \text{drag} \right)$ requires careful tuning of the physical models.

Real-Time Regulatory Logic: Developing the Compliance Engine to run at a high tick rate (e.g., $60\text{Hz}$), capable of enforcing complex rules like virtual "Attack Zones" or energy limits, is computationally demanding and requires a minimal-overhead approach to spatial indexing.

Accomplishments that we're proud of

We are proud of the comprehensive conceptual framework that integrates three distinct engineering domains—advanced physics modeling, competitive AI strategy, and regulatory compliance—into a single, unified system. The emphasis on a compliant leaderboard output addresses a critical need in competitive simulation, moving the tool beyond mere visualization and into actionable race management. The design for the adaptive, multi-layer simulation model represents a robust approach to handling complexity in dynamic mobility systems.

What we learned

Designing the Apex-Gen Simulator taught us the importance of architectural separation in real-time systems. We learned that for competitive simulations, a pure physics engine is insufficient; the competitive outcome is often more sensitive to the agent's decision-making strategy than to minor variations in aerodynamics. Furthermore, we reinforced the understanding that regulatory compliance must be an integrated, first-class citizen in the design, not an afterthought, to ensure the validity of the simulation results.

What's next for Apex-Gen Race & Flow Simulator

The next steps for Apex-Gen would be to move from conceptual design to a working prototype:

Phase 1: Minimum Viable Physics Kernel: Build out the core C++ physics engine for two competing vehicles on a simple oval track.

Phase 2: Implementing the Strategic AI: Develop and integrate a basic Q-learning model for the agents to learn optimal energy deployment and overtaking strategies.

Phase 3: Visualization and Public Leaderboard: Create a Three.js-based web visualization and complete the integration with a database (e.g., PostgreSQL) to persist the compliant, final race results.