People - Sebastian Castillo

Updated CV

The application of engineering and a combination of computational and experimental techniques to improve our understanding of cellular processes, and to develop the ability to re-engineer the cell for human benefit.

Some topics of interest: systems/synthetic biology, robotics, control theory, dynamical systems, complex systems, probabilistic methods,…


Design Principles for Robust Oscillatory Behavior (2010-current)

Here at GMSM we are studying a general three-component enzymatic network model in order to unveil the requirements for robust oscillation. We hope that our findings will help to improve the design of synthetic biomolecular networks and will also have implications in the understanding of natural oscillatory systems.

Some interesting challenges in this project:

  • Performed setup and maintenance of the lab’s computer cluster using Rocks Cluster Distribution.
  • Simulated all the 3284 possible three-component enzymatic oscillators with 300,000 parameter sets each one using python and the numpy, scipy and PyDSTool libraries.
  • Performed data analysis using clustering techniques and visualization tools such as Hive Plots.

Our results will be published soon.

6-DOF Parallel Manipulator (2009-2011)

Originally meant to serve as the base for a Full-Motion Flight Simulator prototype for the Peruvian Air Force (and my undergraduate dissertation at UNI), this 6-DOF Parallel Manipulator was built with the idea of real-time trajectory tracking in mind from the beginning. Additional features include:

  • Point-to-point displacement, using smooth trapezoidal trajectory generators.
  • Logging of position and torque for each actuator during a motion operation.

The manipulator and the controller cards for each motor were built and programmed from scratch.
Some of the toughest challenges in this project:

  • In order to cope with a restricted budget, every component had to be optimally selected, from the structural members, which had to be inexpensive and lightweight, to the electrical motors and the power-stage components. Thus, a great number of inverse dynamical simulations, which compute actuator requirements from any given motion profile, were carried out for each possible combination of components.
  • Dealing with so many actuators was challenging, since no single microcontroller had the resources to run current control algorithms for 6 motors at 20khz, plus a 1khz nonlinear motion control algorithm. Our response to this situation was a distributed control system comprising a dedicated microcontroller (dsPIC) for each motor with a central DSP-based microcontroller (Texas Instruments' C2000 series). The correct behavior of this system depended on sending commands and information precisely at 1khz through an industrial CAN network.

6-DOF Robotic Arm for Automatic Páprika Chili-Pepper Harvesting (2010-2011)

Funded by DIROSE SAC and the Inter-American Development Bank through the Peruvian Goverment's program FINCYT, this project aimed at automating the harvesting process of Páprika chili pepper by combining robotics, artificial intelligence and computer vision.
As part of a team of 6 engineers, I was in charge of control system design, implementation and tests for the six-dregree-of-freedom (6-DOF) robotic manipulator. Entirely built and programmed from scratch, this manipulator now has several capabilities, which include:

  • Point-to-point motion (showed in this video) using joint-space trajectory generators.
  • Command stack (used in this video) for sequentially running several point-to-point operations at different speeds, mixed with end-effector closings and openings.
  • Built-in error handling, in response to events such as crashes, blockages or sensor failures.

Some details about this project:

  • A distributed control system, similar to the one used in the parallel manipulator project, was used.
  • This device had to be made suitable for robustly working long hours in harsh environments. Thus, the robot had to be programed to deal with eventualities, in addition to the command and stack features, and different operating modes. In order to make sense of all this, I adapted software engineering methodologies - regularly used in designing complex non-real-time software projects - to the particularities of multi-threaded real-time control software. This resulted in a modular and scalable piece of software.
  • The strength and vibration analysis of the manipulator’s mechanical structure was performed using Flexible Multibody Dynamics, a technique that analyzes both motion and deformation at the same time without ignoring their coupling. Additionally, it naturally provides a realistic model for direct dynamical simulation, which is required for evaluating the controller’s performance.
  • The arm is -at least theoretically- capable of moving up to 10 times faster than showed in this video. Additional work in non-linear control algorithms and system identification will allow overshoots to be minimized so the arm can go much faster.
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License