CARLA Simulator Developer – Driving Training & Research Platform

We need a developer to build a full-featured simulator on the CARLA open-source platform for practical driving training and human-factors research. The same codebase should support instructor-led training (urban, highway, off-road) and experimental modes for studying stress, drowsiness, and cognitive load on driving behavior. Both left-hand and right-hand driving configurations are required. Core requirements: Ready-made scenarios with dynamic actors (vehicles, pedestrians, cyclists) and weather/time variations, configurable via file or script. Authentic vehicle dynamics (acceleration, braking, steering) using a preferred physics plug-in that integrates with CARLA's API. Research hooks to trigger events (obstacles, reduced visibility, distractions) and log driver inputs, biometrics, and vehicle telemetry at millisecond precision, with a TCP/UDP or REST interface for external device sync (e.g., heart-rate, eye-tracking). A user-friendly training dashboard with admin panel for scenario selection, objective setting, and performance metric review (speed, lane keeping, reaction time). Deliverables: Source code in a private Git repository with clear commit history. A one-click installer or Docker image for Windows and Linux. At least seven demo scenarios per driving context (urban, highway, off-road, hills, night, rain, peak traffic). Documentation: setup guide, API reference, and sample research scripts. A short video walkthrough proving dynamics fidelity and scenario switching. Acceptance criteria: Installer launches CARLA with all assets on a clean machine. Speeds, accelerations, and steering angles match real vehicle data within ±5%. Scenario reload takes greater and equal to 10 seconds without crashing. Logged data streams have timestamp drift greater than 2 ms over a 30-minute session. To apply: Share links to prior CARLA modifications, reinforcement-learning projects, or published human-factors work. All communication and work will remain on Upwork per platform policy. Do not share contact details in your proposal.