1. Trajectory optimization of a 6-DOF arm
Trajectory optmization is an important part of motion/path planning methods used to realize dynamic behaviors for robotic systems. The objective of this project is to realize dynamic behaviors in a 6-DOF robotic arm. An important requirement for this goal is real-time trajectory optimization, i.e., generate new trajectories in the order of milliseconds.
2. Robust stabilizing controllers for hybrid systems
Input-to-state stabilizing controllers can be used to overcome a wide variety of uncertainties in both continuous and hybrid systems. This is demonstrated in the humanoid robot DURUS below:
3. Tree climbing
By employing Partial Hybrid Zero Dynamics we can generate efficient and optimal trajectories for the problem of pole climbing (shown below).
1. DURUS Humanoid
The objective here was to realize human-like walking behavior in the humanoid robot DURUS. DURUS is a 23-DOF robot with 15 actuators.
The goal in this project was to realize dynamically stable running with the planar bipedal robot DURUS-2D. DURUS-2D is a 11-DOF robot with 4 actuators. The running realized in the robot consists of a stance phase, and a flight phase alternating between each other. Input-to-state stabilizing controllers were realized to overcome the phase based disturbances.
AMBER2 is a 7-DOF robot with 6 actuators. This robot is unique due to the presence of feet, allowing for multi-contact locomotion with distinct heel and toe behaviors. Various motion primitives like flat-footed walking, multi-contact walking, standing, lunges, heel-lifts, toe-lifts, swinging, and also bending of knees are demonstrated in the form of dancing below:
AMBER is a 5-DOF robot and has 4 actuators. AMBER used a simple linear control adaptation for tracking trajectories in each joint with the end result being stable walking on the treadmill.
This adaptation was improved further to include motion transitions, with the end result being stable walking on rough terrain.
5. Loop finder
This was my undergraduate project worked in collaboration with Ravi Lanka, in which the robot scans a set of loops and locates the smallest one in the end.