Research in AMBER Lab

The research in AMBER Lab centers ranges from bipedal robotic walking to prosthesis to formal hybrid systems theory to cyber-physical and automotive systems.  Below you will find videos demonstrating results on:
For more videos, see http://www.youtube.com/user/ProfAmes

Experimental Bipedal Robotics

Verifying formal theories on physical robots, and specifically bipedal robots.  Currently, AMBER lab works with the custom built robot AMBER 3M and Cassie (built by agility robotics).  Additionally, the lab works on the custom-built prosthesis, AMPRO, along with a variety of other robotic platforms from hopping robots to flying robots.  In addition, the methods have been applied on a variety of robots through collaborative efforts. 

Nonlinear Control Theory

Through the use of novel Lyapunov functions--rapidly exponentially stable control Lyapunov functions---synthesize real-time optimization based controllers.  These can be combined with safety constraints through control barrier functions.  In particular, control objectives and safety constraints can be realized as constraints in a quadratic program and realized on hardware at over 1 kHz. 

Human-Inspired Robotic Walking

Using human walking data to achieve robotic walking.
  • Begin with human walking
  • Determine constraints from the human data over time
  • Encode these constraints to construct robotic models and controllers
  • Implement these controllers in simulation
  • Obtain "human-inspired" robotic walking experimentally

Multi-contact Hybrid Zero Dynamics

Using human-inspired output functions to achieve robotic walking and, moreover, guarantee that these outputs are invariant through impact.  In addition, the framework of hybrid zero dynamics to multi-contact locomotion, which includes multiple discrete phases of different actuation types. 
Translating robotic walking to obtain stable robust and efficient locomotion on lower-limb prosthesis.  In particular: 
  • Design of novel prosthesis, including: AMPRO, a custom built transfemoral power prosthesis. 
  • Translating robotic walking to prosthesis through the framework of human-inspired control
  • Implementation of advanced nonlinear controllers on prosthesis, including control Lyapunov function based quadratic program based control (CLF-QPs)
Formal methods for general cyber-physical systems, with a special focus on application to automotive systems.  Specifics methods that are being developed include: 
  • A novel form of control barrier function that, when utilized in a quadratic program based feedback controller, allows for probably safe system operation
  • The use of fixed-point and abstraction based formal methods for the synthesis of provably correct controllers.