• Home
  • About
    • Daniel Saromo Mori

      Daniel Saromo Mori

      Mechatronic Engineer & Researcher

    • More info
    • Email
    • LinkedIn
    • Google Scholar
    • Github
    • ORCID
  • Publications
  • Projects

Projects

Auto-Rotating Perceptrons Library

March 22nd, 2021

Keras implementation of the ARP units. To allow an easy use of this perceptrons, the library is available on the Python Package Index (with the command: pip install arpkeras). The Library was presented in a tutorial for the group Papers We Love Guatemala.

GitHub Repository PyPI Project Page Video tutorial (in Spanish)

Mini Tutorial: Deep Reinforcement Learning

September 4th, 2020

I have been a teacher in the Specialization Diploma on Development of Artificial Intelligence Applications, offered by the PUCP Graduate School. As a presentation of the AI for Games course I taught, I made a mini tutorial about Deep Reinforcement Learning.

Video

Robot learning using DDQN and NEAT for a 2 DOF pointer robot

May 1st, 2020

Just for fun, I wanted to build a robot that learns by experience. I designed and built a small robot that learns to control a laser beam using Deep Reinforcement Learning, Neuroevolution, and Computer Vision. The 2 DOF robot learned to point its laser beam to reach a target located at the center of two marks. It received the graphic information from a smartphone’s camera. Then, it identified the beam and marks position in real-time using computer vision. The high-level processing is done on a computer and the Arduino acts as the robot low-level controller. For the communication between these devices, I used my PyDuino Bridge Library, which is freely available for the community.

The algorithms tested were the Double Deep Q-Learning (DDQN) and NeuroEvolution of Augmenting Topologies (NEAT). A better performance was obtained using the first algorithm, whose results are shown below.

First image: Physical setup of the robot while executing its first trials (please ignore the little hole in my wall, as the robot does). Second image: Robot learning to move the beam to the desired position (marked with a blue dot).

PyDuino Bridge Library

April 30th, 2020

Library for transparent bi-directional communication between Python and Arduino. Available on the official Arduino Library Manager and on the Python Package Index (with the command: pip install pyduinobridge).

GitHub Repository PyPI Project Page

My 8 DOF spider robot: making it learn to walk

November 7th, 2018

When I was studying machine learning, I wanted to use those techniques in a physical environment. Thus, I built a 8 DOF spider robot. I followed the Kamrani’s rapid prototyping methodology: start with a simple design, build it, add more features, make the needed corrections, and then iterate the process several times.

Iterative process of the spider design. I started with a simple walking robot, but it could not move itself without falling down. After that, I added two DOF at each leg, and increased the torax's size. Then, I increased the number of legs, and added one rotative distance sensor. In the next step, I simplified the leg design, but it was a bit unstable. Hence, for the last iteration, I used a 1 DOF four-bar linkage mechanism at the end of each of the four legs. The lateral rotation of each limb is achieved with the servomotor that fixes the leg to the torax. After the iterations, I ended up with a functional spider robot.

For the mechanical structure, I laser-cutted MDF and then assembled the pieces.

First image: CNC machine making the mechanical parts of the spider. Second image: Building of the spider's torax.

Regarding the electronics of the spider, it has a 9 DOF IMU (for measuring the 3D orientation), it has a current sensor (to measure the amperage consumed during the gait), and other components.

Finished spider robot.

For the control algorithm, I developed an algorithm to maximize a reward function without applying reinforcement learning techniques, but supervised ML. The algorithm I proposed to make the robot learn to walk was later presented as part of my bachelor’s thesis.

Kalman filtering with Arduino and MATLAB

September 2nd, 2018

I implemented two Kalman filters using MATLAB.The first Kalman filter combined a noisy measurement and a model-based prediction to find the optimal estimate of the signal.

The output of a know mathematical model received additive noise. I used a Kalman filter to estimate a signal (in red), that is more similar to the original one (in blue) than the noisy signal (in black).

The second Kalman filter blended three synthetic noisy signals, working as a sensor fusion Kalman filter.

There were three synthetic signals (in yellow, cyan and magenta), which I Kalman filtered in order to obtain the red signal. I observed that the estimated signal is more similar to the clean signal than any of the noisy signals.

The signal-to-noise ratio (SNR) was used to quantify the amount of noise present in the signals.

Then, I used four accelerometer signals to have a more reliable measurement by applying real-time sensor fusion with Arduino.

The four accelerometers used in the sensor fusion (in order of appearance at the image): the accelerometer embedded in the Intel Curie chip of the Arduino 101, two ADXL335 modules, and a GY85 module. The Kalman filter was implemented in Arduino.

BB8-shaped robot

March 19th, 2016

I built a robot whose shape was inspired from the BB8 robot that appeared in a Star Wars movie (Episode VII). It is a rolling robot with a free-moving domed head on the top, magnetically kept in place during the robot locomotion.

The robot has two main parts: the internal structure and the external sphere. First, I made the internal structure base by laser-cutting MDF. Then, I added the motors, the electronics, and the battery. Later, I used a servomotor to attach the piece whose magnets allowed the robot to turn its head and keep it in place while the body is rolling.

First image: Bottom part of the robot's internal structure. Second image: Inner components of the robot inside the opened sphere.

After that, I put the internal structure inside the spherical body.

Gaze of the robot's inside.

Finally, I closed the sphere and added the domed head. This piece has magnets and casters on its bottom side.

First image: My BB8 robot with its sphere closed and its head in place. Second image: Original BB8 robot from the movie Star Wars Episode VII.