Overcoming Initial Convergence in Multi-objective Evolution of Robot Control and Morphology Using a Two-Phase Approach

Proceedings of the Genetic and Evolutionary Computation Conference

Martín

Samuelsen

et al. 2018

Self Cite

For robots to handle the numerous factors that can affect them in the real world, they must adapt to changes and unexpected events. Evolutionary robotics tries to solve some of these issues by automatically optimizing a robot for a specific environment. Most of the research in this field, however, uses simplified representations of the robotic system in software simulations. The large gap between performance in simulation and the real world makes it challenging to transfer the resulting robots to the real world. In this paper, we apply real world multi-objective evolutionary optimization to optimize both control and morphology of a four-legged mammal-inspired robot. We change the supply voltage of the system, reducing the available torque and speed of all joints, and study how this affects both the fitness, as well as the morphology and control of the solutions. In addition to demonstrating that this realworld evolutionary scheme for morphology and control is indeed feasible with relatively few evaluations, we show that evolution under the different hardware limitations results in comparable performance for low and moderate speeds, and that the search achieves this by adapting both the control and the morphology of the robot.

Section: Evolution Of Robot Morphologymentioning

confidence: 99%

Real-world evolution adapts robot morphology and control to hardware limitations

Proceedings of the Genetic and Evolutionary Computation Conference

Martín

Samuelsen

et al. 2018

Self Cite

“…The simplest approach for the evolution of morphology for real-world robots is to evolve the design in simulation, then manufacture a select few individuals and verify them in hardware. There are several examples of this being done for legged robots [68,80], as well as more unconventional robot designs [31,49]. The challenge with this approach is that the evolved individuals suffer from the effects of the reality gap, which can be even more substantial when both control and morphology are optimized.…”

Section: Evolution Of Morphologymentioning

confidence: 99%

Multi-objective evolution of fast and stable gaits on a physical quadruped robotic platform

2016 IEEE Symposium Series on Computational Intelligence (SSCI)

Tørresen

Glette

2016

Self Cite

Robots are used in increasingly complex environments and need to be able to adapt to changes and unexpected events. This has traditionally been solved by changing the control of a robot, but having an adjustable body can unlock new and powerful adaptive capabilities. An adaptive morphology allows tuning of the physical structure of the robot to different, often conflicting, dynamic requirements, including speed, stability, and efficiency. It can also unlock new functionalities that might not be possible with static morphologies, including variable gearing and multiple locomotion modalities. Even with the potential benefits of morphological adaptation, the methods and technology are still not at a point where there is wide-spread use of adaptive morphologies in physical robots.The main goal of the thesis is to develop methods and technology to enable adaptation of the physical body of a robot to new real-world environments. An evolutionary approach is taken, and to what degree evolutionary algorithms are able to exploit the dynamic morphology of a legged robot is investigated. The feasibility of continuous adaptation of morphology in realistic outdoor environments is also explored.A quadruped mammal-inspired robot with the ability to continuously adjust the length of its legs during operation has been designed and implemented as part of the work outlined in the thesis. Evolutionary algorithms are used to optimize both the control and morphology of the robot to different hardware conditions and walking surfaces in the lab. To achieve this, a new gait controller concept with an adjustable complexity is introduced. This allows evolution in scenarios with a wide range of evaluation budgets. A final proof-of-concept implementation of adaptive morphology is also demonstrated. Our robot was shown to be able to adapt its body continuously while walking in different unstructured outdoor terrains, significantly outperforming a non-adaptive approach.The thesis concludes that adaptation of the physical body of a robot is feasible, and in fact, already shows significant benefits with current technology and methods. Evolutionary algorithms are shown to be effective for adaptation of morphology in a range of different conditions. By developing new methods and technology, as well as demonstrating their utility through real-world experiments, we hope to inspire others to use adaptive morphology on their physical robots.

“…The field of evolutionary robotics uses techniques from evolutionary computation to optimize control and-less oftenmorphology. Evolution of both control and morphology together is usually performed in simulation and presents additional challenges due to the complex search-space [13]. The difference between performance in a simulator and a real-world counterpart is referred to as the reality gap, and often makes it very challenging to transfer a result to the real world.…”

Section: Introductionmentioning

confidence: 99%

Self-Modifying Morphology Experiments with DyRET: Dynamic Robot for Embodied Testing

2019 International Conference on Robotics and Automation (ICRA)

Martín

Tørresen

et al. 2019

Self Cite

If robots are to become ubiquitous, they will need to be able to adapt to complex and dynamic environments. Robots that can adapt their bodies while deployed might be flexible and robust enough to meet this challenge. Previous work on dynamic robot morphology has focused on simulation, combining simple modules, or switching between locomotion modes. Here, we present an alternative approach: a self-reconfigurable morphology that allows a single four-legged robot to actively adapt the length of its legs to different environments. We report the design of our robot, as well as the results of a study that verifies the performance impact of self-reconfiguration. This study compares three different control and morphology pairs under different levels of servo supply voltage in the lab. We also performed preliminary tests in different uncontrolled outdoor environments to see if changes to the external environment supports our findings in the lab. Our results show better performance with an adaptable body, lending evidence to the value of self-reconfiguration for quadruped robots.