|Ph.D Student||Spitz Jonathan|
|Subject||Bio-Inspired Controllers for Dynamic Locomotion|
|Department||Department of Mechanical Engineering||Supervisor||Professor Miriam Zacksenhouse|
|Full Thesis text|
In recent years there has been a growing interest in the field of dynamic walking and biologically inspired robots. However, while walking and running on a flat surface have been studied extensively, walking dynamically over terrains with varying slope remains a challenge.
A biologically inspired feed forward controller was designed for a simulated compass biped model. The biomimetic controller, inspired by Central Pattern Generators, coordinates the timing of torque pulses and achieves stable dynamic gaits over a limited range of slopes, between -1.2 and .8 degrees. Applying short torque pulses, instead of enforcing joint trajectories, enables the controller to achieve energy efficient gaits by exploiting the system's natural dynamics.
The robustness to slope variations of the feed forward controller is vastly increased by adding a once per cycle feedback. The terrain's slope is measured and used to adapt the feed forward controller to the terrain by modifying both the controller's frequency and the torque amplitude once per step. Relying on this minimal feedback strategy, the controller successfully traverses terrains with slopes ranging from -8 to degrees, comparable to most slopes found in human constructed environments. The eigenvalues of the linearized Poincare map were numerically computed for each slope to verify the system's orbital stability.
The controller's parameters were optimized using a multi-objective genetic algorithm. The algorithm evolved populations of controllers by evaluating their performance in walking speed, energy efficiency, rate of convergence to a limit cycle and robustness to slope variations. The algorithm generated controllers with distinct abilities which can be roughly divided into four groups: (I) fast walkers, (II) efficient walkers, (III) quickly converging walkers and (IV) good climbers.
Based on the same concepts, a bio-inspired crab-walking framework was designed for the humanoid robot ATLAS that, like humans, can take advantage of quadruped gaits to traverse steep slopes. Cleverly designed gaits provide robustness to rough terrain without requiring extensive feedback. The controller includes forward and backward crawling patterns, rotation patterns, and sit-down and recovery sequences. The latter are activated autonomously if the robot detects that it tipped over. The performance and robustness of each locomotion pattern were investigated over a wide range of slopes. Crab-walking is especially adept at crawling forward on steep downward slopes (up to -54 degrees) and crawling backward on steep upward slopes (up to degrees). These quadruped gaits enable ATLAS to complete the dismounted mobility part of the DARPA Virtual Robotics Challenge which includes crossing a mud pit, a hilled area and an area with debris.