Some of my Work

An intelligent Neuro-Dynamic Walking Engine for Humanoid Robots

Using tools of artificial intelligence, mainly artificial neural networks, this work presents an intelligent walking engine for humanoid robots. This engine uses dynamic neural networks with intelligent feedback for gait generation, a modified Zhang neural network for a singularity-robust inverse kinematics solver, and feedforward neural networks for neuro-adaptive control. The Atlas humanoid robot model in simulation is used to test and verify the capabilities of the neuro-dynamic walking engine. Results show that the engine is capable of generating ZMP stable walking gaits and executing them using the Atlas robot in simulation.

SMIRFF: Smart Maintenance, Inspection, and Repair Free-Flyer

SMIRFF is a concept for a compact, intelligent robotic free-flyer that possesses capabilities which allow operators to inspect, maintain, and minimally repair external components of space systems. The concept was developed for the 2014 NASA/NIA RASC-AL competition, where it was awarded First Place in the graduate division. Having a cubic shape and using a cold-gas thruster based propulsion system, SMIRFF is capable of maneuvering in any direction and at any orientation to arrive at a required location. Due to the complexity of space structures and the need to interact with such structures for inspection and maintenance, SMIRFF possesses two robotic arms with a novel moving base concept. Combined with it’s on-board vision and inspection sensors, SMIRFF is truly the future of space structure inspection.


Developing Unmanned Ground Vehicles

  • Hardware integration
  • Control system design
  • Sensor Fusion
  • Communication system
  • Software development

IMG_20150214_155204         IMG_20150212_083759


Neural Network Control of a Humanoid Robot

NN-blockdiag atlas

  • Work was done to compete in the DARPA Robotics Challenge (DRC)
  • The work was based on developing a walking controller that allows the Boston Dynamics ATLAS humanoid to walk
  • Using the ZMP criterion, a trajectory for the feet and robot center of mass are generated
  • Using a full-body inverse kinematics solver, the required joint angles or stable walking are genertaed
  • The robot nonlinear dynamics are learned on-line using a feed-forward, multi-layer, neural network
  • A control system that utilizes the “learned” dynamics is used to track the desired joint angles for stable walking


Adaptive optimal control of a 2-DOF helicopter

  • Solving the optimal control LQR problem online, by learning the controller gains
  • The scheme is based on adaptive critics and alternates between evaluating the control policy and updating it based on the system performance
  • Similar to how the human brain works
  • Scheme is applied to a 2-degree-of-freedom helicopter experimental setup


Airship optimal control and wind vector estimation for navigation

  • Airships: vehicles that utilize gases which are less dense than air (helium) to generate buoyancy force to keep aloft
  • Airship dynamics are substantially different from conventional aircraft dynamics
  • Airships have high sensitivity to wind disturbance

  • Two main guidance laws were developed: i) Track-specific law (TS) ii) Proportional navigation (PN)



Design of an Automatic Landing System for a Small UAV

land1 land2

land4 test


Vision based navigation of a quadrotor

  • Develop software for a vision-based control system applied the AR.Drone quadrotor using python
  • ROS (Robot Operating System) provides libraries and tools to help software developers create robot applications (BSD license).
  • OpenCV (Open Source Computer Vision) is a library of programming functions for real time computer vision (BSD license).

block3  block4