Portfolio
Smart e‑Puck Car: Vision‑Based Pedestrian & Obstacle Detection for Safer Line‑Following Robots
A Webots-based autonomy prototype combining line following, OpenCV perception, auditable safety logic, and supervisor-driven validation with reproducible tests and rich telemetry.
Project details
What we delivered
Overview
An end-to-end autonomy stack on an e‑Puck robot in Webots: line following + vision detection + distance reasoning + deterministic safety decisions + full logging.
Focus: measurable, explainable behavior in controlled scenarios (simulation-validated).
What We Built
Continuous line following control loop.
OpenCV-based classification (pedestrian vs obstacle) from forward camera frames.
Safety policy with explicit thresholds and state transitions (cruise → slow → stop / avoid).
Supervisor controller for scenario control, measurement, and reproducible logging.
Pipeline
Sensors (camera + proximity + odometry) → line error → vision detection → distance estimation → safety policy → motor commands → supervisor logging.
Logged Metrics
Detection confidence per frame
Decision-to-actuation latency
Minimum distance to pedestrian/obstacle
Avoidance success and collision events
Line-following deviation over time
Tech Stack
Webots (R2025a)
e‑Puck simulation model
Python
OpenCV
Webots Supervisor APIs