🪛 A Few of the PINN Use Cases

Here's a non exhaustive list of some of the possible PINN use cases across different domains.

🧬 PINNs are great for proprioception and state estimation tasks, such as:

• Legged robots discovering ground friction on unknown terrain
• Manipulators discovering payload mass and inertia when grasping unknown objects
• Soft robots estimating material properties as they deform and age
• Sensor failure compensation
• Correcting the DVL drift in underwater systems

💡 MPC Dynamics Surrogates for robotics:

• Real-time control of soft robots
• Excavator bucket control adapting to unknown soil
• Surgical tool control adapting to unknown tissue stiffness
• Quadruped gait adaptation across terrain types

🌀 Addressing the Sim-to-Real Gap in robotics:

• Discovering the actual friction, damping, stiffness that the simulator got wrong
• Correcting MuJoCo/Isaac Sim dynamics from a few real-world runs
• Transfer learning from sim to real via parameter discovery

🌊 In ocean and environmental applications:

• Ocean current reconstruction from sparse sensor data
• Coral reef monitoring
• Pollutant transport tracking
• Tidal pattern decomposition

🔋 For construction:

• Soil parameter discovery from excavator sensor data (friction, cohesion, moisture)
• Crane load dynamics with unknown wind forces
• Terrain classification through dynamics
• Structural health monitoring

🧪 In the surgical and medical domains:

• Tissue stiffness estimation during surgery from force-torque data
• Real-time deformation tracking for AR surgical overlay
• Patient-specific tissue modeling from sparse pre-operative imaging
• Catheter navigation, i.e. discovering vessel wall properties in real-time

📐 In structural and civil engineering:

• Bridge health monitoring for discovering structural degradation from vibration sensors
• Building response to earthquakes for discovering actual damping and stiffness
• Wind load estimation on structures from sparse accelerometer data
• Crack detection through parameter change detection

🔥 In the energy sector:

• Battery state of health for discovering internal resistance and capacity degradation
• Wind turbine performance for discovering aerodynamic parameters from SCADA data
• Solar panel degradation detection
• Power grid fault detection through parameter anomalies

📡 In aerospace:

• Flight dynamics parameter estimation from sparse flight test data
• Engine health monitoring for discovering compressor degradation
• Aerodynamic coefficient estimation from flight data
• Re-entry vehicle thermal protection for discovering ablation rates

🌍 And lastly for geoscience purposes:

• Seismic inversion for discovering subsurface material properties from surface measurements
• Groundwater flow for discovering aquifer permeability from sparse well data
• Volcanic activity monitoring for discovering magma properties from seismic data
• Climate model parameter estimation