Design and Summary Analysis Draft 3
MEC 1281
Summary_Analysis
Draft #3
By Chin Yong Sheng
21st Feb 2021According to the article, "Soft robotic arm..." (2020), researchers
from MIT constructed a soft robot that is capable of estimating its 3D
configuration through its skin that is covered with sensors. However,
the sensors and neural network still lack refinement to accurately
capture subtle and dynamic motions.
There are two
advantages the soft robot has over traditional rigid robots. Due to its soft body, it has the ability
to move in infinite number of ways at any time. Secondly, it does not
rely on motion-capture cameras to provide feedback regarding 3D movement
and positions. Instead, it relies on its own sensors. "The sensors can
be fabricated using off-the shelf materials"(Truby,n.d,Para
5). Dr.Truby commented that the sensors are easy to fabricate, meaning
any lab will be able to construct their owns sensor. Dr.Truby also
highlighted that the materials used for constructing the sensor should
possess "piezoresistive" properties. The material's resistance varies
according to the motion. The change in resistance affects voltage output
linearly. The resulting voltage output will then be relayed to the robot.
However, its ability to move without restriction results in increased difficulty to use it in "control" applications. To alleviate this problem, new robots have to undergo deep learning. Doing so, the robot will progressively be more effective in capturing meaningful feedback signals. I hope to explore more feasible options such as utilizing better sensors and to provide better deep learning methods to reduce training time.
While
the soft robotic arm is more suitable than traditional robot in
real-life scenarios, technology has yet to be refined.“ Think of your own body: You can close your
eyes and reconstruct the world based on feedback from your
skin,”(Rus,n.d,Para 7). One future aim is to construct a
soft robot that is dexterous, and capable of manipulating objects in its
surrounding through better sensors and deep learning.
Incorporating a better material for future prototypes should be looked into as the initial material used is not ideal. The material used couldn't stretch much without making cuts on the material itself. In the article "Advanced carbon materials...", it states that carbon materials such as Carbon Nanomaterial and Graphene are strong contenders for this particular application that we are interested in. It is said that these 2 materials can be easily assembled into multi-scaled macroscopic structures, resulting in good flexibility and conductivity. The article further states that carbon materials can be prepared through different approaches, thus creating even more room for future development.
Secondly, an improvement to the deep learning
method can be desirable. Traditionally, a process called Simultaneous
Localisation and Mapping (SLAM) is used to map environment using a
machine's sensors. However, the process is incapable of mapping a
dynamically changing environment due to uncertainties. Filters are added
to the sensor to counteract this undesirable behaviour. A filter known
as Smooth Variable Structure Filter(SVSF) is used to provide a better
estimate to the uncertainties and errors captured by the sensors. In the
article "Robust SVSF-SLAM for Unmanned Vehicle in Unknown Environment"
the graph shown in figure 6 illustrates how the filter works. Through
estimation of the errors, the filter changes the estimated trajectory
consistently. Through filtering, the estimated trajectory is shown to be
very close to the system's true trajectory.
Dr.Truby
mentioned:" As hypothesized, the
sensors did capture the trunk's general movement,
but it was really noisy. Substituting the original sensor material for
carbon-based material will be a great improvement in "noise" reduction,
thus leading to a shorter learning time for the robot. Coupled with an
improve learning method, an ideal soft robot might just be possible in
the near future.
References:
Zhang.Y, Jian.M(2017, Sep 1) Advanced Carbon Materials For Flexible And Wearable Sensors. 60(11), 1026-1062.
https://engine.scichina.com/publisher/scp/journal/SCMs/60/11/10.1007/s40843-017-9077-x?slug=fulltext
Matheson,R.(2020, February 16). Soft robotic arms uses flexible sensors to understand its position. Control Engineering. https://www.controleng.com/articles/soft-robotic-arm-uses-flexible-sensors-to-understand-its-position/
Institute of Informatics and Applications, University of Girona(2008). The SLAM Problem: A survey. http://eia.udg.es/~qsalvi/papers/2008-CCIAa.pdf
Denim.F, Nemra.A, Louadj.K(2016). Robust SVSF-SLAM for Unmanned Vehicle In Unknown Environment. 49(21), 386-394.
Comments
Post a Comment