Research Projects
Our lab's research spans across various domains, ranging from hardware development to the creation of controllers utilizing Optimal Controls, learning-based controls, and safety-critical controls. The following projects serve as a testament to this diverse spectrum of work.
Active Projects
Learning a formally verified CBF in stochastic environment
Neural Network Based CBF with completeness guarantees
Dynamic Manipulator
A highly dynamic, torque controllable manipulator for quadruped robots.
Force control for Robust Quadruped Locomotion: A Linear Policy Approach
A framework for designing controllers to achieve robust blind quadrupedal walking using force control thorugh learnt linear policies.
Energy Based Control Barrier Functions
A new class of CBFs for robotic systems that augment kinetic energy with the traditional forms.
Completed Projects
Learning Based Control of Stoch BiRo 1
Design and Control of a custom made low cost bipedal robot - Stoch BiRo
Collision Cone Control Barrier Functions (C3BF)
Control Barrier Functions for Kinematic Obstacle Avoidance :A Collision Cone Approach
Off-Policy Average Reward Actor-Critic with Deterministic Policy Search
Policy gradient theorem for average reward criteria with deterministic policy.
Dynamic Mirror Descent MPC for Model-based Model-free Reinforcement Learning
A framework for utilizing experience for generating predictive simulations and learning from them.
Learning Controllers for Robust Bipedal Locomotion on Challenging Terrains
A framework for sythesizing controllers to achieve blind bipedal walking on challenging terrains thorugh learnt linear policies.
Robust Quadrupedal Locomotion on Sloped Terrains
This paper presents a linear policy approach to achieve walking on sloped terrains
Stoch 1
A complete description of the hardware design and control architecture of our custom built quadruped robot, called the Stoch
Active Spine on Stoch 2
Learning Active Spine Behaviors for Dynamic and Efficient Locomotion
Trajectory Optimization on Stoch 2
Trajectory based Deep Policy Search for Quadrupedal Walking
Augmented Random Search (RL Algorithm)
Reinforcement Learning using ARS (Augmented random Search)to generate Gaits