Traxium

every lap smarter, every decision sharper

Comment

Inspiration

We were captivated by the high-stakes strategic decisions in competitive mobility — from Formula E teams managing energy consumption to logistics companies optimizing delivery routes. Yet we noticed a gap: existing tools were either overly complex physics simulators or simplistic models that couldn’t capture the nuanced trade-offs between speed, efficiency, and strategy. We envisioned a platform that would democratize strategic analysis, allowing anyone to explore "what-if" scenarios in competitive mobility systems through an intuitive, visual interface.

What it does

  • Traxium is a next-generation, lightweight, and interactive mobility simulator that merges AI, strategy, and performance analytics to transform how teams plan, test, and optimize competition. It models multiple moving agents — such as race cars, drones, or delivery bots — under real-world constraints, allowing users to run “what-if” scenarios and see how changes in rules, weather, or strategy affect results in real time.
  • Teams can fine-tune variables like speed limits, tire wear, fuel, weather, or pit strategies and instantly visualize how every choice impacts performance. Traxium serves as a virtual race lab, turning raw data into actionable race-day intelligence.
  • Powered by reinforcement learning and predictive models, its agents learn to overtake efficiently, conserve energy, and adapt to evolving race conditions. The AI Strategy Coach analyzes previous runs to suggest optimal pit stops and energy use, while the AI Race Manager acts as a virtual strategist — providing live tactical insights such as, > “Driver A is pitting — stay out for two more laps to gain position,” or “Rain incoming — delay pit and switch to intermediates later.” ### Key Features:
  • Multi-agent simulation of cars, drones, or delivery bots with configurable profiles
  • Constraint-based testing with energy limits, weather conditions, and strategic parameters
  • Live leaderboard tracking positions, gaps, and key metrics in real-time
  • Side-by-side scenario comparison with synchronized playback and analytics
  • Interactive visualization with map views, agent trails, and event markers
  • Export and replay capabilities for deeper analysis and sharing — ###User Interface
  • The user interface combines clarity and control through a live 3D map, dynamic dashboards, and synchronized simulation playback. Users can experiment, compare, and optimize decisions seamlessly in one environment.
  • For racing teams and mobility innovators, Traxium is more than a simulator — it’s a strategic intelligence platform. It helps teams make faster, smarter, and data-driven decisions while training AI systems that learn like real drivers.
  • Traxium turns simulation into intelligent strategy — where every lap, every call, and every decision gets smarter.

How we built it

Architecture & Technologies

  • Core Simulation Engine
  • Backend Framework: Python 3.11+ with SimPy for discrete-event simulation
  • Simulation Core: Custom priority queue-based event management system
  • Optimization Engine: Google OR-Tools with CP-SAT solver for strategic constraint optimization
  • Numerical Computing: NumPy + Pandas for performance calculations and analytics
  • Machine Learning: Scikit-learn for predictive modeling of agent behavior

Intelligent Agent System

  • Agent Architecture: Object-oriented design with hierarchical state machines
  • Behavior Models: – Kinematic Bicycle Model for vehicle dynamics – Monte Carlo Tree Search (MCTS) for tactical decision-making – Reinforcement Learning (PPO) for adaptive strategy optimization
  • Memory Management: Redis for real-time agent state persistence
  • Decision Engine: Custom rule-based system with fuzzy logic controllers

Optimization & AI Systems

  • Constraint Optimization: Google OR-Tools CP-SAT for: – Pit Stop Strategy Optimization – Energy Management Scheduling – Overtaking Sequence Planning
  • Predictive Models:
  • XGBoost for tire wear prediction
  • LSTM Networks for lap time forecasting
  • Transformer Models for opponent behavior prediction
  • Multi-objective Optimization: NSGA-II for Pareto-optimal strategy discovery

Frontend Dashboard: React + D3.js for real-time visualization

Simulation Approach

We used probabilistic models rather than heavy physics, with performance calculated as:

performance = base_profile × f(energy_level, conditions) + strategic_adjustment + ε

This lightweight approach enabled rapid iteration while maintaining meaningful strategic depth.

Challenges we ran into

  • Synchronization Complexity: Achieving perfect parallel alignment across simulations using deterministic event scheduling and shared random seeds.
  • Performance vs. Realism: Balancing computational efficiency and realism through a probabilistic event system.
  • Intuitive Configuration: Translating high-level strategic questions (like “What if battery capacity was reduced by 15%?”) into precise parameters.

- Visualization Clarity: Developing synchronized timelines and comparative heat maps to simplify multi-agent visualizations.

Accomplishments that we're proud of

  • Built a functional comparative scenario engine that synchronizes two complex simulations perfectly
  • Developed a modular architecture adaptable across domains — from motorsports to logistics
  • Achieved sub-second response times for complex scenario comparisons
  • Designed an intuitive UI accessible to non-technical users

- Built a robust data model for real-time metrics and post-session analytics

What we learned

Technical Insights

  • Discrete-event simulation offers massive performance advantages for strategic modeling
  • Deterministic random number generation ensures reproducibility
  • Agent-based modeling captures emergent system behavior effectively ### Domain Understanding
  • Energy constraints reshape competitive dynamics across all mobility domains
  • The physics–strategy relationship forms rich strategic landscapes

- Visual comparison outperforms numerical analysis in understanding complexity

What’s next for Traxium

Enhanced Simulation Capabilities

  • Integration with real-world data APIs (weather, traffic, energy pricing)
  • Machine learning agents that develop and optimize their own strategies
  • Multi-stage scenario modeling for endurance races or multi-day logistics ### Platform Expansion
  • Collaborative features for team-based strategy design
  • Plugin architecture for domain-specific modules (aerodynamics, traffic patterns, etc.)
  • Mobile app for real-time strategy adjustments during live events ### Commercial Applications
  • Specialized editions for motorsports, logistics, and urban planning
  • Educational version for engineering and business strategy courses

- Esports integration as a backend for competitive strategy games

Summary

Traxium represents not just a tool — but a new paradigm for understanding and optimizing competitive mobility systems through the power of comparative simulation.

Built With

Share this project:

Updates