Engineering

Code, simulation and autonomous systems

Engineering work presented with the problem, the method, the tests and the limits. Every figure on these pages comes from a retained result file or a public repository.

01Featured project

Autonomous systems · Simulation · Programming

A Python controller that drives a TORCS race car around the Corkscrew circuit with no human input. It plans a racing line and a speed profile from the track geometry, follows them with feedback, and was tuned against the simulator’s own lap time, from a first completed lap of 279.22 s to a locally validated 65.968 s with the final controller, P1.

Read the case study →

Best lap
65.968s, validated
Damage
0every counted run
Repository runs
5/5identical result
Source
Publicrelease branch, commit e71b6fe
Measured speed along the validated lap · 65.968 s lap, 313 recorded samples

02Inside the case study

What the case study covers

Eleven sections, from the challenge to the lessons, with measured telemetry, architecture, test runs and a link to the source.

  1. 01Overview and challenge
  2. 02Simulator and development setup
  3. 03Controller architecture
  4. 04Steering and path following
  5. 05Speed control, braking and the measured lap
  6. 06Eight stages of development
  7. 07Lap-time progression
  8. 08Testing and reliability
  9. 09Final demonstration
  10. 10Source code and evidence
  11. 11Lessons from the lap

03What it demonstrates

Disciplines in the evidence

  • Evidenced

    Programming and automation

    A real-time Python controller, a one-command validation script and checksummed run tooling, in a public repository.

  • Evidenced

    Simulation

    Closed-loop testing in TORCS, and a search that used the simulator itself as an exact objective.

  • Evidenced

    AI and autonomous systems

    An autonomous racing controller. It is deterministic, model-based control rather than a trained network, and was developed with AI coding assistants.

04Software and tools

Verified tools

Only tools that appear in the project’s files are listed.

Tools and where they are evidenced
ToolUsed forEvidence
Python 3.9.13Controller and test runnerValidation run reports
NumPy 2.0.2Plan arrays and numerical modelRepository requirements file
TORCS and the SCR serverSimulator and client interface (UDP port 3001)Repository README and run logs
gym_torcs / Snake Oil clientStarter client loop, used unmodifiedRepository third-party notices (MIT)
CMA-ESPer-window parameter search against the official lap timeRepository README (method; library not recorded)
Git and GitHubVersion control and public sourcePublic repository
IBM Bob, Codex, Claude CodeAI coding assistants, credited by stage in the project historyRepository README disclosure; project history reports