Physics-based models bridge neuroscience and robotics! 🧠 Researchers from @CarnegieMellon University from NeuroMechatronics Lab started with a biomechanics model initially developed at West Virginia University to understand how coordinated muscle forces produce movement. The model examines forces and mechanics at the mechanical level, how neurons activate muscles, muscles generate forces, and forces move limbs. They used @MathWorks @MATLAB's Simscape Multibody to build and simulate these mechanical systems. Extended the approach to lower limbs. Investigated motor pathologies. Asked how biological neural networks organize themselves to perform sensorimotor computations, leading to work on Artificial Physics Engines. The human body is a mechanical system governed by physics, but movement is controlled through interacting layers. Neurons coordinate muscles. Muscles generate forces. Forces move limbs. To understand movement, you need to model the entire chain from neural activity to motion. Applied the same thinking to robotic control. Developed new neural network architectures for predicting and controlling motion in complex robotic systems. Physics-based simulations using Simscape Multibody train and evaluate these networks, providing the mechanical environments needed to test how neural controllers actually move real systems. EMG (electromyography) captures muscle activity signals. Decode those signals and you get intuitive control interfaces. A participant with spinal cord injury can use EMG from their arms to control a virtual hand, visualizing intended movement even when the limbs won't respond. The same approach extends to exoskeletal systems. EMG-based remote control. Movement intention becomes action through a device instead of paralyzed limbs. 🔗 That’s an awesome project, read more about it here: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →