Realization of multi-objective evolved continuum robots using 3D printing
Continuum robots are recognized as one of the most flexible and versatile mobile robots that are capable of performing various kinds of motions to navigate in unknown and challenging environments. However, the large number of degrees of freedom leads to the difficulty in designing a continuum rob...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2016
|
| Subjects: | |
| Online Access: | https://eprints.ums.edu.my/id/eprint/12016/1/Realization%20of%20multi-objective.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Continuum robots are recognized as one of the most flexible and versatile mobile
robots that are capable of performing various kinds of motions to navigate in
unknown and challenging environments. However, the large number of degrees of
freedom leads to the difficulty in designing a continuum robot. Moreover, an open-ended
synthesis problem arises whereby there exists no formal models thus far for
a designer to determine the optimum control strategy, body structure, number of
segments and suitable segment lengths during the design stage. Additionally,
conventional methods for designing continuum robots do not consider the
optimization of multiple objectives. As such, there has not been any research
carried out thus far on co-evolving both the morphology and controller of
continuum robots using a multi-objective evolutionary optimization approach.
Therefore, in this research work, a system is developed to automatically design and
optimize both the morphology and controller of continuum robots by employing a
novel hybridized Genetic Programming and self-adaptive Differential Evolution
algorithm. A multi-objective evolutionary algorithm is incorporated into the artificial
evolutionary optimization process to simultaneously maximize the locomotion
performance and minimize the complexity of the continuum robots. In addition, a
novel GP tree-based encoding structure is proposed to allow for the representation
of the continuum robot's morphology and controller to be optimized simultaneously
during co-evolution. The artificial co-evolutionary process is carried out by using the
Webots physics simulation software. Two types of continuum robots are to be
evolved in this research, namely the snake-like continuum robot (SLCR) and multi-branching
continuum robot (MBCR). The outcome of this work shows that the
Pareto-optimal front of evolved solutions are successfully obtained for the simulated
SLCRs where the evolved heterogeneous SLCRs can perform lateral undulation,
narrow path crawling, vertical undulation and lateral rolling moving behaviours for
locomotion. Additionally, the evolved solutions of the MBCRs are converging to a
point where the MBCR with the least number of segments turns out to be the
dominating solution. In order to validate the simulated results, the evolved SLCRs
are transferred to real world for physical testing using 3D printing technology. The
physical testing results demonstrate that the evolved SLCRs can be successfully
transferred from simulation to the real world for actual physical deployment in its
task environment. An 82.55% transference accuracy is achieved in this work which
demonstrates that the proposed multi-objective co-evolutionary algorithm is
feasible and practical to be employed for the automatic design of continuum robots |
|---|