𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗔𝗻 𝗔𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀 𝗥𝗮𝗰𝗶𝗻𝗴 𝗔𝗴𝗲𝗻𝘁 𝗶𝗻 𝗧𝗢𝗥𝗖𝗦

I joined the IBM global AI racing competition.

I use TORCS for testing.

My goal is a clean lap on the Corkscrew track. No crashes. Then I will make it fast.

The plan has three steps:

  • Rule-Based Control: Use PID for steering and braking.
  • Machine Learning: Use PyTorch to replace hardcoded rules.
  • Optimization: Push physics limits.

My tools:

  • Simulator: TORCS.
  • Language: Python with snakeoil3.
  • Local AI: ibm-granite-3.2-8b-instruct via LM Studio.

I will share these details:

  • AI prompts.
  • Generated code.
  • Control system math.
  • Debugging logs.

Follow for robotics, control theory, and Python updates.

The first log covers steering and braking logic soon.

Source: https://dev.to/dglezz/building-an-autonomous-racing-agent-in-torcs-19gm Optional learning community: https://t.me/GyaanSetuAi