Motivation Having a complete picture of the ground interactions of a legged robot is a necessity in enabling high- speed and dynamic ground locomotion. Current sensing methods are inadequate to address the unique demands of robotic legged locomotion, and are vulnerable to inertial noise upon high acceleration. M=70 kg V=4.5 m/s Using new design principles and methodologies, we have developed a low cost, robust footpad sensor designed for running robots. This approach maps the local sampling of pressure inside a polymeric footpad to forces in three axes using machine learning. The foot sensor is a monolithic composite structure that is composed of a piezoresistive sensor array PCB completely embedded in a protective polyurethane rubber layer. This composite architecture allows for compliance and traction during ground contact, while deformation alters the measured stress distribution. Large normal forces of 424N are measured in the Z- axis with a normalized RMSE of 1.2%, and for the shear forces, the range is 233N with a normalized RMSE of 10.1% in the Y-axis, and 219N with a normalized RMSE of 8.3% in the X-axis. Least Squares Artificial Neural Network (LSANN) is a new approach to reduce the computational time for convergence to obtain a useful estimator for normal and shear forces by 29.2%. Another area of research is material modelling and FEA simulation to better inform sensor placement. Design Principles Improvements Approach Results Currently force sensing shoes are being developed to help assist the elderly and disabled for slip prediction, fall prevention and mitigation. Athletes can also benefit from the real-time in-situ force data collected to better optimize their training workouts. Future Directions Stress Distribution under Shear Automated CNC mill data collection setup 2nd generation MIT Cheetah footpad MIT Cheetah jumping on grass while untethered MIT Biomimetic Robotics Lab website: biomimetics.mit.edu Michael Chuah: [email protected] [email protected] www.michaelchuah.me