Titulo Estágio
RaceCrewAI/RaceEngineerAI: Integrating an AI Race Strategist for Enhanced Racing Strategy
Local do Estágio
Rua Pedro Nunes, 3030-199 Coimbra
Enquadramento
Racing has evolved into a highly competitive domain, demanding sophisticated strategies for optimal performance. Quicktrend is exploring advancements in this area through its projects RaceCrewAI and RaceEngineerAI. This proposal focuses on addressing the need for dynamic, data-driven decision-making by developing a functional prototype of an "AI Strategist." The project aims to demonstrate the integration of machine learning and reinforcement learning techniques to build a tool capable of predicting lap times, modelling tyre degradation, suggesting pit-stop strategies, and offering basic real-time tactical advice. The intern will work on the RaceCrewAI/RaceEngineerAI projects, with access to relevant data (e.g., telemetry, race logs). The core idea is to create a functional prototype that can analyse race scenarios and provide actionable strategic advice.
Objetivo
The primary objective of this proposal is to design, develop, and evaluate a prototype of an "AI Strategist", demonstrating the integration of three core modules to support strategic decision-making:
• Lap Time Prediction (Module A: Foundational Model):
o Develop a foundational model to predict lap times based on core car telemetry data (e.g., speed, laps on tyres), setup parameters, fuel status, and potentially weather conditions.
• Tyre Degradation Modelling & Pit-Stop Optimisation (Module B: Core System):
o Create a simplified model to estimate tyre wear and its impact on performance, based on stint history and available tyre state proxies.
• Develop a pit-stop suggestion module using inputs from the foundational Module A and the tyre state estimator.
• Reinforcement Learning Agent (Module C: Basic Implementation):
o Design a basic RL agent for illustrative dynamic race decisions, considering a simplified state space (e.g., current lap, basic tyre status, fuel level).
o Integrate the agent to conceptually use features from Modules A and B and decide on an action.
The overall goal is to produce a functional prototype that demonstrates the concept of unifying prediction, wear estimation, and adaptive learning to offer insightful dynamic strategic suggestions for racers.
Plano de Trabalhos - Semestre 1
• F1 – State-of-the-Art Analysis, Literature Review, and Foundational Research (30% of semester):
o Conduct a focused review of existing techniques for lap time prediction, tyre degradation modelling, pit stop optimisation, and applications of Reinforcement Learning in racing.
o Critically analyse existing academic papers, open-source projects, and relevant functionalities within RaceCrewAI/RaceEngineerAI to inform pragmatic choices.
Evaluate the suitability of various straightforward machine learning models, optimisation heuristics, and basic RL frameworks for rapid prototyping.
Key Outcome: Based on this research, select promising and feasible algorithmic approaches and architectural patterns for the core implementation of each of the three modules (A, B, C.
• F2 – Core Requirements Elicitation & System Architecture Design (30% of semester):
o Define core functional requirements for each module and the integrated AI Strategist, informed by F1.
o Specify essential data requirements (focusing on core car telemetry from simulators like iRacing/Assetto Corsa), input/output formats for the initial modules.
o Design the overall modular architecture for the AI Strategist prototype, outlining interactions between modules and data flow.
• F3 – Prototyping Module A (Foundational Lap Time Prediction) (40% of semester):
o Data collection, cleaning, and preprocessing of essential telemetry data.
o Develop and implement the initial version of the lap time prediction model, using the approach selected in F1.
o Conduct initial feature engineering, model training, and basic validation of this initial model.
Plano de Trabalhos - Semestre 2
• F4 – Core Module Development and Integration (70% of semester):
o Module A (Foundational Model) Finalisation: Refine and complete the lap time prediction foundational model.
o Module B (Core System) Development: Implement the simplified tyre life estimator and pit-stop suggestion module. Integrate the core Module B with outputs from the foundational Module A.
o Module C (Basic Implementation) Development: Design and implement the basic RL agent. Train the agent in its simplified environment.
o Initial System Integration: Integrate functional versions of Modules A, B, and C.
• F5 – System Testing and Conceptual Evaluation (10% of semester):
o Full System Integration & Testing: Integrate all three developed modules into the AI Strategist prototype.
o Conceptual Evaluation: Evaluate the integrated prototype against defined functional metrics.
• F6 – Prototype Finalisation, Documentation, and Thesis Report (20% of semester):
o Prepare essential documentation.
o Write the final master's thesis report detailing the research, prototype development process, results of the conceptual evaluation, and conclusions.
Condições
This is a non-remunerated internship. The intern will have full access to Quicktrend's computational resources, including servers and relevant software. Access to simulator accounts (e.g., iRacing, Assetto Corsa) and their respective data streams/APIs will be provided as necessary for the project's development and testing phases. The company will offer guidance from experienced personnel and a conducive environment for research and development at its IPN incubator facilities.
Observações
It is understood that the nature of research and development, even for initial systems, may lead to adjustments in the project plan and specific module implementations. The proposed plan serves as a guideline, and flexibility will be maintained to adapt to findings and opportunities that arise during the development process, ensuring the project remains focused on delivering a functional system.
Orientador
Mário Gil Figueiredo Abrantes
gilabrantes@quicktrend.eu 📩