This is a guest post. Thank you Fabrice R. Noreils.
1. The Objective
The main objective is to design a medium size humanoid robot (around 90 cm) and create a niche market which is considering humanoid robot like an art. The robot will be able to walk like a human, thanks to the innovative design of mechanical structure and associated algorithms, perform task and, very important point, equipped with outifts/outshells created by famous designers in order to meet different communities’ expectations.
2. How to reach it
In order to reach this ambitious target, several objectives have been defined:
- The first objective is to create innovative walking gaits – closest as possible to human ones, i.e. no more bending knees – real heel strike – active use of an articulated forefoot. These gaits will be more or less “hardcoded“. The main goals are in one hand to demonstrate the feasibility of such gaits and on the other hand to understand how the different parts of the body are coordinated and how to take benefit from the dynamic of the robot.
- The second objective is to design a controller, based on the knowledge gained from the previous work, which will be connected to sensors and takes into account the dynamic of the robot. It means that the walking gaits will not be hardcoded anymore but generated online. It will lead to faster and smoother walking gaits as well as the possibility to react to unexpected events/obstacles.
- Once the mobility is achieved, it can be possible to add more sensing capabilities like a camera, different kinds of gripping tools and/or specific accessories in order to achieve new tasks.
- The last objective is more oriented towards art and design. To please the audience, mobility and instilled the audience that the robot is “more or less clever” is not enough, it must be beautiful. To achieve this goal, the idea is to bring famous designers that will be able to create outfits/outshells matching different kind of communities’ expectations (fashion, mangas, mecha…).
The first objective of this project is under completion and in order to show the progress, mainly focusing on the walking gaits, I published videos that are available on YouTube.
Two different gaits are considered: straight walk gait and Turning gait. For each one there are several videos (Video 1 is the oldest, video n is the newest).
odoi Straight Walk Video 2 with speedx4
Odoi straight walk Video 3 speedx4
The first video regarding the turning gait:
More videos will be added from time to time and technical details can be found on the blog I have created: Artbot.
4. The robot
This section provides a detailed description of both the mechanical structure as well as the hardware of the actual robot.
Figure 1 gives an overview of the current robot. It is 75cm tall.
4.1 Mechanical structure
The robot is equipped with
- An articulated feet;
- An articulated pelvic;
- An articulated torso;
Most of the brackets are made of resin (and printed with a Form 1+) except the Pelvic where aluminum has been required. Indeed because of the robot weight, the previous brackets and Pelvic structure made of resin bent leading to discrepancies between real angles and theoretical ones at the hip that were too important.
The foot which is an important element of the robot is detailed on Figure 2.
The main benefits of the mechanical design are:
- Innovative gait – no more bending knees;
- Walking gait closer to human gait;
- Omni-directional walking;
- Save energy as the robot is not bending knees which is a real benefit when the robot runs on
- Possibility to change the stride length;
- Participation of the whole body.
So far, from my knowledge only two hobbyists/researchers were able to achieve a walking gait without bending knees:
- Masahiko Yamaguchi, nickname is Dr Guero, featured a modified KHR-3HV which was able to walk on a floor almost like a human does . However the robot is not able to turn and I do not know if it is possible to change the stride length and get the same astonishing result.
- Tomotaka Takahashi with his latest creations, Robi, Kirobo/Mirata and more recently Robohon, developed a patented walking gait without bending knees  – However these cute robots are equipped with a quite large footprint in order to maintain stability.
The hardware – see Figure 3– will be composed of:
- An OPEN CM9 board which is connected to the Dynamixel servos from Robotis. The robot is equipped with 2 * MX106T, 4*MX64T, 11*MX28T and 9 AX12 servos;
- A PIXY Cam to do basic object/color recognition;
- A 9DOF Razor IMU which will be used to control the balance;
- FSR sensors;
- Murata Rotary potentiometers to measure the heel orientation;
- A Raspberry Pi 2/3 will be added soon in order to generate gaits (and more) online.
In the current version of the robot, only the OPEN CM9 is used and it is connected to the Dynamixel servos. It will be connected to the FSR and Murata sensors very soon.
The Raspberry will be added in the coming months. It will be then connected to the OpenCM9, the Pixy CAM and the 9DOF Razor IMU.
5. How to do
The principle of the approach is depicted on Figure 4.
Each walking gaits is decomposed into phases:
- Lateralization on the right side;
- Left leg swing;
- Lateralization on the left side;
- Right leg swing.
For each phase, a sequence of movements is created. Each sequence is composed of a start (T0) a duration (D), an objective () and a template curve to follow.
The phases are the input of the kinematic simulator which generates the movements of each limbs. This software takes into account the kinematics of the robot, it computes also the CoG (Center of Gravity). The output is a set of angular positions at different timestamps for each limb.
In order to compute the angular position and associated speed for each servo, another software is used, referred to as the “commands generator”.
The inputs are the angular position for each limb (processed by the Kinematic simulator) and the duration of each phases. The possibility to change the duration of each phase gives us the possibility to play (a bit) with the dynamics. If the duration is too short, the velocity at a given time for some actuators will be too important and it will generate oscillations leading to instabilities.
The output of the “commands generator” is a table of commands that will be used by the OpenCM9 board. If a list of commands have to be sent at a given time t, then a pointer to a list of (servoId, position and speed) is created – see Figure 5.
6. Artistic design
I think that Artistic Design is really very important if one want to introduce robots in the human environment that can be accepted and/or tolerated by the population. One step further will be the development of “artistic robots” that can be considered as piece of Art. So far only the Japanese robotics community is really addressing this topic.
This is why I am eager to work with designers in order to create “outer shell” (or even outfits) that can fit the robot skeleton. I initiated a collaboration with Dacosta Bayley, a Canadian artist, who is running among other things MarchOfRobot on Instagram (every day in March, artists are pushing sketches picturing robots – see #marchofrobots2016). He ran successfully a kickstarter campaign in 2014 in order to publish a book about his work.
7. Next steps
The next steps will consists in bringing all the software on the raspberry in order to generate “on the fly” the walking gaits. Sensors will be connected to the processing boards in order to adjust the gaits to the reality of the terrain. Dynamics will be introduced step by step.
Carry on a project like this one alone is very challenge, therefore if there are some people interesting to participate with skills in software programming, mathematics, electronics or artistic, please do not hesitate to contact me.