IMU Sensor Fusion & Kinematics Developer: Real-Time 3D Motion Visualizer

We are building a wearable boxing-motion prototype using a 5-node IMU network (ESP32-C3). We have a stable data pipeline streaming at 100Hz, but we need an expert to bridge the gap between raw data and a physically defensible 3D pose visualization. The Challenge: Our system uses one IMU per distal limb segment. We explicitly acknowledge the observability problem: with only 5 nodes, limb positions are underdetermined. This is a constrained inference problem, not a simple measurement task. We need an expert who can use anatomical constraints to infer plausible elbow and knee positions during high-velocity boxing strikes. What You Will Do: Audit our GitHub Repository: Work primarily on the "madgwick" branch. Robust Sensor Fusion: Improve robustness of orientation estimation under dynamic motion and reduce drift to enhance stability. Constraint-Based IK: Implement an Inverse Kinematics (IK) solver that uses biomechanical assumptions to compute endpoint positions from a single distal IMU node. Offline Validation: Use our existing Pose Logging system (JSON Lines) to validate algorithms against recorded movement scripts. Deliverables: Improved live pose stability (reduction in lag and "visual explosions"). Reproducible validation against recorded logs. A clear technical statement of what the 5-node system can and cannot realistically estimate. Required Skills :Expertise in IMU Sensor Fusion (Madgwick, Mahony, or Kalman). Deep mastery of Quaternions and coordinate transforms. Experience with IK and skeletal pose logic. Proficiency in Python (NumPy, SciPy, Matplotlib).