Project: Finalize real-time joystick controller (PID/MPC tuning, smooth tracking, stability) We have an application that is nearly complete. The remaining work is the joystick control layer: converting tracking error into smooth, stable analog stick commands (RX/RY) in real time. This is NOT a computer vision model training job and NOT a machine learning task. Detection/tracking and state estimation are already implemented. The work is primarily control engineering + signal shaping + real-time tuning. What you’ll do Improve RX/RY command generation for fast acquisition and smooth steady tracking Reduce jitter, stick-slip, overshoot, stalls, and inconsistent corrections under noise/latency Tune and/or redesign: error normalization & scaling response shaping (curves, tanh/logistic, piecewise gain scheduling) deadzone modeling and compensation rate / acceleration limiting and anti-windup handling Ensure state/authority gating is consistent end-to-end (what the controller computes is what is sent) Handle transitions cleanly: target changes / reacquire measurement dropouts and prediction/grace windows mode changes (acquire vs track) Iterate based on logs + real hardware testing until behavior is stable and “human-smooth” Control methods (expected familiarity) We already have a working pipeline and are open to improvements using: PID (with anti-windup, derivative filtering) MPC (optional, if you have strong experience—focus is real-time practicality) Multi-rate control concepts (e.g., estimator update < control tick, latency compensation) Robustness against measurement dropouts and delay Required Strong experience with analog stick/joystick control: shaping, smoothing, deadzones, limiters Solid control intuition (stability vs responsiveness tradeoffs) Strong Python engineering (real-time loop constraints, profiling, clean logging) Ability to debug from detailed logs and implement focused fixes quickly Hardware to test the controller end-to-end on a real device input path (details shared privately) Deliverables Final tuned joystick control pipeline (acquire + track modes) Stable behavior across lock/prediction grace states and target transitions Clean, documented parameters and a short tuning guide (“how to adjust for device feel”) Verified test results with logs demonstrating stability and smoothness Notes We will provide the full codebase, logs, and current controller pipeline. We’re looking for someone who can ramp up quickly and deliver improvements through focused iteration and testing