Eric Nauli Sihite
PhD in Mechanical Engineering from the University of California San Diego, with research focus in robotics system design, dynamic modeling, control, and estimation.
Objective: seeking for a full time position in robotics research, design, and controls.
Quick links:
updated 5/2021
Resume
Curriculum Vitae (academic)
Linkedin profile
GitHub profile
Micro ball-balancing robot:
IROS 2015 paper
ICRA 2019 paper
Patent
Video
Drive/coast motor modeling:
CASE 2019 paper
Video
MIP state estimations:
ACC 2018 paper
Research Experience
Northeastern University
(Postdoctoral Researcher)
2020 – 2021
PI: Prof. Alireza Ramezani
Aerobat
Aerobat – a flapping wing robot with dynamic wing morphing capability, inspired from bats.
Lead the research project for developing a small and lightweight ornithopter which mimics bat flapping wing behavior. I developed the novel flapping wing mechanism which incorporates wing folding and expansion during upstroke and downstroke respectively. This achieves a more efficient flapping gait by minimizing negative lift during the upstroke.
I designed the CAD model of the mechanism using Solidworks and fabricated the components using 3D printer and a high resolution laser cutter which is capable of cutting carbon fiber plates for the mechanisms. This robot was extensively tested in front of our airflow generator, load cell, motion capture camera, and high speed camera. I also process the data collected in these experiments for performance evaluation.
Additionally, we studied and developed a simulation model in Matlab for evaluating the feasibility of this robot by modeling the unsteady aerodynamic forces acting on the wings using lifting line theory and Wagner’s function. Using this model, we proposed a controller framework which utilizes four degrees-of-freedom (plunge, mediolateral movement, elbow flexion-extension, and feathering) to stabilize the robot’s orientation to perform an upside-down perching maneuver. We also developed a different control framework which utilizes a change in morphology to influence the resulting flapping gait and aerodynamic forces acting on the body.
The mechanical design of the armwing structure is published in R-AL (IROS 2020) while the controller frameworks are published in CDC 2020 and ACC 2021.
Paper links:
R-AL 2020
CDC 2020
ACC 2021
Video links (featured in the lab’s website):
video 1
video 2
video 3
Optimization-free Controller Framework using Reference Governor
Harpy – a thruster-assisted bipedal robot where the thrusters can be used to stabilize frontal dynamics and adjust the ground reaction forces to satisfy friction constraints.
The bipedal robot locomotion on a difficult to navigate terrains can be a very difficult task, even humans which have great dexterity and sense of balance may have difficulty in walking on slippery or uneven surfaces. In our lab, we are developing a thruster-assisted bipedal robot called Harpy where the thrusters can be used to assist the robot in navigating difficult terrains and exhibit multi-modal locomotion by simply jump over these obstacles.
In this project, I developed a simulator for a quadripedal and a thruster-assisted bipedal robots in Matlab. In both systems, I developed a Reference Governor-based controller framework to enforce ground reaction force constraints. The Reference Governor works by adjusting the controller references to satisfy some constraints, which in this case is the ground friction cone constraints, while keeping this manipulated references as close as possible to the desired references. This optimization-free method allows us to enforce constraints online as long as we have an accurate model of the robot’s and actuators’ dynamics.
In the Harpy’s case , the addition of thruster on the body allows the adjustment of the ground reaction forces to prevent slips as it walks on uneven terrain or slippery surfaces. Additionally, we showed the multi-modal capability of our robot in the simulation by jumping over several obstacles using the thrusters.
Paper links:
ACC 2021
University of California San Diego
(PhD Graduate Student Researcher)
2012 – 2019
PI: Prof. Thomas Bewley
Dissertation:
Mechanical Design, Dynamic Modeling, State Estimation, and Feedback Control of a Micro-Ball-Balancing Robot at High Yaw Rates.
Committee chair: Prof. Thomas Bewley, University of California San Diego.
Micro Ball-balancing Robot (MBBR) [2013 – 2019]
The most recent Micro Ball-Balancing Robot prototype.
A small scale robot which balances itself on top of a ball using three omni-directional wheels. This robot is one of the smallest ball-balancing robot in the world and uses low-cost components, with weight of 650 g, 25 cm tall, and costs less than $200 to build. This project is a very challenging control problem due to the miniature size, low-cost components, high noise, and the coupled nonlinear 3D dynamics inherent to the system.
This is my dissertation project and there are currently 2 publications on this project: the design paper and the high-yaw rate dynamic model and estimation paper. The mechanical design also features the patented “midlatitude and orthogonal” omniwheel placement which reduces the wheel slip and wheel coupling interference.
The design paper focused on the mechanical design, omniwheel placement, component selection and the simple linear controller using Successive Loop Closure (SLC) technique. The second paper focused on the nonlinear high yaw-rate dynamic modeling and state estimation for the MBBR using Extended Kalman Filter (EKF). This is done so that the state estimation is accurate and can achieve a more stable balancing during high yaw-rate maneuvers. The robot has undergone several design improvement since the first paper which includes the optimized double-row omniwheel design, DC motor, motor driver and gearbox selection.
Finally, the most recent and the best controller and state estimation algorithms for the MBBR are presented in the dissertation. A controller which takes advantage of the system’s inherent non-minimum phase dynamics is presented and can achieve very good stabilization even under high yaw rates.
The design paper is published in IROS 2015:
paper link and video link.
The modeling and estimation paper is published in ICRA 2019:
paper link and video link
The control algorithm that can drive and spin fast at the same time:
video link.
Coasting Brushed DC Motor Driver Model Identification [2017 – 2019]
Two primary brushed DC motor driving modes: drive/brake and drive/coast. The current during each drive, brake, or coast are shown above.
The PWM driven, bi-directional brushed DC motor drivers are separated into two different driving modes depending on how the MOSFETs inside the H-Bridge are configured during the LOW PWM signal as shown above:
- Drive/brake = either the low or high side MOSFETs are closed while the opposite sides are opened.
- Drive/coast = all MOSFETs are opened which drains the motor current through the flyback diodes.
The commonly used linear PWM motor torque model is only valid for drive/brake mode while the drive/coast mode exhibits a nonlinear behavior that varies depending on the motor inductance and the driving PWM frequency. This nonlinear behavior can cause significant control problem if not compensated or ignored.
Model verification setup using the mini-Dynamometer developed in our lab and a Linux computer called Beaglebone Black (BBB).
In this project, we derived and identified the drive/coast motor dynamic model, verify the model accuracy under a dynamometer, and propose a real-time implementation strategy in an actual robot. The coasting motor model was derived using Mathematica, which then verified by using a mini-Dynamometer developed in our lab. The dynamometer collect the input and output data from a simple motor – flywheel system which was used to verify that the nonlinear motor torque behavior can be determined using the new coasting model we derived.
Once the drive/coast model’s accuracy has been verified, we used the model to control a Mobile Inverted Pendulum (MIP) robot with drive/coast motor driver in real-time. The controller designed for drive/brake driver was transformed to the equivalent PWM duty for the drive/coast driver. The compensated controller has the same control performance as the drive/brake motor driver, showing that the compensation worked for real-time control applications.
This work is published in CASE 2019:
paper link, video link 1 and video link 2.
Video 1 shows a MIP that is controlled using a controller designed for a drive/brake driver. The compensated drive/coast controller achieved a similar performance, which indicates that the compensated motor torque is approximately identical to the drive/brake’s. Video 2 shows the MIP driving around with drive/coast motor drivers.
Attitude Estimations of a Mobile Inverted Pendulum (MIP) using High Yaw Rate Dynamic Model [2017 – 2018]
The educational Mobile Inverted Pendulum (eduMIP) developed in our lab, with motion capture markers attached.
Attitude estimation on a Mobile Inverted Pendulum (MIP) robot using several estimators which accuracy were compared to the motion capture data. This research was done as a prove of concept of the high yaw-rate dynamic modeling and state estimation for the MBBR project. The MIP robot has similar inverted pendulum dynamics as the MBBR but much more simple which makes experimenting the methodology on this robot a logical intermediate step for the MBBR.
The estimators tested were Complementary Filter, Complementary Kalman Filter, linear model Kalman Filter and high yaw-rate dynamic model Extended Kalman Filter (EKF). The nonlinear dynamic model for the MIP was derived which models the effect of a nontrivial yaw rate. The derivations were derived symbolically using Wolfram Mathematica. This model was then implemented in an EKF and the estimation accuracy was compared with the motion capture system. We also compared the accuracy of the more common attitude estimators listed above under high yaw-rate.
This work is published in ACC 2018:
paper link.
University of Michigan Ann Arbor
Optical Fiber Alignment [2011 – 2012]
Research done in University of Michigan Ann Arbor with Prof. Kenn Oldham on an optical fiber alignment system using a mechanically controlled piezoelectric actuator which can be realigned by using a heat plate and a thermal glue. The thermal glue melts when heated up by the heat plate and the fiber is aligned on the glue, which then sets in as the glue cools down. This process can be repeated if the fiber needs to be realigned as many times as needed.
I was responsible for the initial research, preliminary test setup assembly and data gathering. I left to begin my doctoral study in UC San Diego in the middle of the project, so this project was completed by someone else.
This work is published in ACC 2013:
paper link.
Publications
Quick links to the publications shown above.
2021
P. Dangol, A. Lessieur, E. Sihite, and A. Ramezani, “A HZD-based Framework for the Real-time, Optimization-free Enforcement of Gait Feasibility Constraints,” International Conference on Humanoid Robots (Humanoids), 2021, in press.
E. Sihite, P. Dangol and A. Ramezani, “Unilateral Ground Contact Force Regulations in Thruster-Assisted Legged Locomotion,” Advanced Intelligent Mechatronics (AIM), 2021, in press.
A. Ramezani, P. Dangol, E. Sihite, A. Lessieur, and P. Kelly, “Generative Design of NU’s Husky Carbon: A Morpho-Functional, Legged-Aerial Robot,” International Conference on Robotics and Automation (ICRA), 2021. [link]
E. Sihite, A. Darabi, P. Dangol, A. Lessieur, and A. Ramezani, “An Integrated Mechanical Intelligence and Control Approach Towards Flight Control of Aerobat,” American Control Conference (ACC), 2021. [link]
K. Liang, E. Sihite, P. Dangol, A. Lessieur, and A. Ramezani, “Rough-Terrain Locomotion and Unilateral Contact Force Regulations with a Multi-Modal Legged Robot,” American Control Conference (ACC), 2021. [link]
A. Lessieur, E. Sihite, P. Dangol, A. Singhal, and A. Ramezani, “Mechanical Design and Fabrication of a Kinetic Sculpture with Application to Bio-Inspired Drone Design,” Unmanned Systems Technology XXIII (UST), 2021. [link]
E. Sihite, A. Lessieur, P. Dangol, A. Singhal, and A. Ramezani, “Orientation Stabilization in a Bio-Inspired Bat-Robot Using Integrated Mechanical Intelligence and Control,” Unmanned Systems Technology XXIII (UST), 2021. [link]
2020
E. Sihite and A. Ramezani, “Enforcing nonholonomic constraints in Aerobat, a roosting flapping wing model,” Conference on Decision and Control (CDC), 2020. [link]
E. Sihite, P. Kelly, and A. Ramezani, “Computational Structure Design of a Bio-inspired Armwing Mechanism,” IEEE Robotics and Automation Letters (RA-L), 5(4), pp. 5929-5936, IEEE, 2020. [link]
2019
E. Sihite, D. Yang, and T. Bewley, “Modeling and state estimation of a Micro Ball-balancing Robot using a high yaw-rate dynamic model and an Extended Kalman Filter,” International Conference on Robotics and Automation (ICRA), pp. 8577-8583, IEEE, 2019. [link]
E. Sihite, D. Yang, and T. Bewley, “Derivation of a new drive/coast motor driver model for real-time brushed DC motor control, and validation on a MIP robot,” International Conference on Automation Science and Engineering (CASE), pp. 1099-1105, IEEE, 2019. [link]
2018
E. Sihite, and T. Bewley, “Modeling and state estimation of a mobile inverted pendulum robot using Extended Kalman Filter under high yaw rate,” American Control Conference (ACC), pp. 5831-5836, IEEE, 2018. [link]
2015
D. Yang, E. Sihite, J. Friesen, and T. Bewley, “Design and controls of a micro ball-balancing robot (MBBR) with orthogonal midlatitude omniwheel placement,” International Conference on Intelligent Robots and Systems (IROS), pp. 4098-4104, IEEE, 2015. [link]
Co-authored with my lab partner, Daniel Yang.
2013
E. Sihite, Z. Qiu, and K. Oldham, “Modeling and control of optical fiber micropositioning in a thermal adhesive,” American Control Conference (ACC), pp. 6263-6268, IEEE, 2013. [link]
I designed and built the test setup for the fiber optic alignment experiment and performed the preliminary experiments. I am credited as a contributor to the paper but did not write the paper.
Technical Skills
Skill list:
- Robotics electro-mechanical design.
- Linear systems theory.
- Optimal controls and estimation (LQR, Kalman Filtering).
- System dynamic modeling (e.g. Lagrangian dynamics formulation).
- Computer vision (stereo vision, feature detection).
- Sensor fusion.
- Path planning (A*, RRT).
- Model predictive controls.
- Adaptive controls.
- System identification.
- Numerical simulation.
- CAD modeling.
- Linux embedded systems.
Programming Languages: C, C++, Python, Matlab.
Programs used: Matlab, Mathematica, Solidworks, LaTeX, Eagle, MS Office.
Education
University of California San Diego, La Jolla, California, USA
PhD, Engineering Science (Mechanical Engineering – 2019
University of Michigan Ann Arbor, Ann Arbor, MI, USA
MSE, Mechanical Engineering – 2010
BSE, Mechanical Engineering – 2008
Other Skills
Languages: English, Indonesian, basic Japanese.
Other Info
Birthday: March 17, 1987.
Hobbies: building robots, badminton, cooking, bicycling, video games, manga, anime, memes, reddit.
Eric Nauli Sihite's Personal Webpage 