Cellulo V2 Mechanics
The changes in the mechanics were mainly fueled by a need to improve the locomotion system as well as to introduce modularity while maintaining the design criteria defined in the main wiki of V2 (cost, scalability, usability, production...). The decisions are taken and the results are detailed below.
The changes applied in the current version are the following:
- Introduction of 3 modules; Main module with locomotion and core electronics, battery module, and User interface module.
- The connections between the modules are done mechanically through a rail and assisted by magnets to provide an additional feeling of connection plus robustness.
- 2 Versions of locomotion are designed. Magnetic drive and Omniwheels.
- Omniwheels are in-house designed and tested.
The biggest burden spotted in the previous version involved the locomotion system. In particular, Cellulo is based on a novel technology that enables the robot to move holonomically. A holonomic movement is such that each of the degrees of freedom is directly controllable. When using wheeled robots, this tends to be a problem, since the wheels can only be driven in one direction, and that is the reason why the Omni wheels or Mecano wheels were created since they are active in the rotational direction. Still, thanks to the small rotational parts in the radius of the wheel, it doesn't block the displacement perpendicular to the rotation. Seeking the same effect, the CHILI team came up with the idea to use magnets to drive the wheels. In a detailed manner, the wheel consists of a metallic ball that is magnetically paired with the shaft of the motor, and when the motor moves it drives the wheel, however, if the ball wants to rotate in a direction that is not the same as the magnet direction, the shear force would not be sufficient to interrupt the movement.
More details about the design can be found in the thesis Cellulo: Tangible Haptic Swarm Robots for Learning.
This design presents multiple advantages, however, it comes with some inconveniences. First of all, the size of the robot requires small parts, hence, the bearings and the magnet are parts that are harder to find out and accommodate in the design, i.e. the bearings are 3mm in diameter, nylon made and allocated directly in the chassis, hence, to provide adequate functionality the tolerances in the bearing hole must have small tolerances and smoothness, which is hard to produce. As for the ball, the initial design was done with metallic balls of 14 mm and a casting of 1mm to provide friction in the movement. Those balls are custom-made, and hence an appropriate supplier must be found, the last supplier used can be found in the following email: 1007785188[at]qq.com. Be aware that it is a Chinese supplier and has limitations in the language and payment methods. Additionally, these design incurs a maintenance cost. Due to the nature of the movement and the design, dust, and particles tend to be accumulated in the bearings, which difficult the move over time. The same movement and friction between the floor, the magnets, and the balls might cause erosion in the coating surface, which deteriorates the movement. Because of those reasons, it is required to do maintenance periodically, and parts substitution.
Alternatively, the use of omniwheels is proposed. This provides trades-off against the ball wheels. On one hand, the omniwheels reduce the number of parts that need to be obtained for the locomotion, simplify the design, and provide locomotion slightly better (subject to the quality of the omniwheels). On the other hand, the production of omniwheels of the initially proposed diameter (14 mm), has had to be done in-house, which is difficult for its scalability and requires specialized tools. Additionally, the use of omniwheels requires some adequations in the design. Either the kinematics of the robot is changed to use the same motors, or it is required to find motors with a 90º shaft that satisfy the requirements in r.p.m. and torque. Ultimately, and depending on its use, the omniwheels could be allocated outside of the robot, which was an initial proposal in version 1.
The chassis has been designed following the same principles as version 1, which is intended to be graspable, portable, and robust. The design is slightly bigger in size to accommodate the modularity requirements.
Version 2 has 3 modules, a main module, a battery module, and a user interface module. The connection between each of the modules is provided by a sliding grove with enough tolerance and thickness to withstand its uses (The parameters of thickness and tolerance must be adjusted depending on the material of the chassis and further testing of usability). Additionally, magnets have been placed in the inside face of each of the contact surfaces between modules to provide a locking force keeping the modules in place, providing a better feeling and reliability in the mechanism. It is possible to experiment with a small mechanism to add friction or locking forces.
The outside chassis has an initial version "Main Module V1.step" which focuses on proving the concept. Further, a second version was developed thinking more about aesthetics and usability, ending up in two options: Main Module V2A.step and Main Module V2B.step, both of them equally accepted internally in the development company. It is advised to do user testing to evaluate and gather more feedback, as well as iterate in the design.
The chassis can be allocated regardless of the locomotion system that is used (With the exception of the version with omniwheels outside). Four versions of the locomotion were developed:
- Magnetic ball drive.
- Inside omniwheels.
- Inside omniwheels at 90º.
- Outside omniwheels.
As for the magnetic ball, the first thing to be taken into consideration is that it requires minimum production setup. The highest challenge of the magnetic ball drive is to produce a chassis capable of allocating the bearings (3mm nylon) and the balls with enough tolerances to hold the bearings and the balls and not excessive friction to prevent locomotion. The last version produced was printed in a resin printer by formlabs Form 2. The initial developments were done in an extrusion low-end printer, which was not sufficient to provide the quality necessary to allow the displacement of the robot in a sustainable way. However, it is necessary to remark that version 1 was printed in extrusion mode. There are companies providing 3d printing services which would be of adequate quality for this version, however, it might require some testing of the materials, tolerances, and thickness. The version can be found in the following link: Magnet Balls Base.step.
The version with omniwheels is a much more simple base, which simplifies the production, increases robustness and reduces costs. However, it still requires some tuning to improve the omniwheels. In combination with a new version of electronics where there is only one PCB, as well as bigger omniwheels, it is expected to provide better results, as well as facilitate the production of omniwheels. This version, Celluo V2.1 OmniWheels v6.step, consists of a version using the same motors as per the initial version 1 (pololu motors), however, due to maintaining the angles of the motors it is required to change the matrix of locomotion in the firmware.
Using motors of 90 degrees, Inside Omniwheels 90D.stp, it is possible to maintain the same kinematics for the locomotion. It is also easier to allocate the space of the wheels and the motors, however, the motors so far tested do not provide the pair torque/speed provided by the other options. Further research is necessary for the models.
In the first iteration of Cellulo V1, it was compared the performance between outside omniwheels and the magnetic drive balls, ending up with similar results, with slight vibrations for the omniwheels, although slightly better trajectory tracking. This design was tested with the new version of Cellulo, and the base can be found here Outside Omniwheel Base.step.
The magnetic ball drive was successfully migrated to the new version. It is recommended to test depending on the materials and production method. The bottom lower lid could be improved, with a geometric design that absorbs better the bendings and incrusts rates itself with the other lid.
The rail for the modularity works smoothly with the material tested, but it is by no means a perfect solution, hence it is possible to improve it.
It is required to do usability and likeability tests for the last designs with real users based on the use cases.
The Omniwheels is believed by the author to be a better solution in all aspects, including costs, time and effort to produce and mount, robustness in the long term, and performance, however, it still requires some iteration, finding a better size with the new electronics and finding a supplier or a production system that guarantees the performance.
Additionally, the motors could be improved, finding a better-suited torque/rpm solution, especially for the 90º motors.