Exploring the Challenges of Implementing Evolved Robots in Hardware

Researchers from various universities have published a technical paper titled “Practical hardware for evolvable robots,” which delves into the challenges of implementing evolved robots in hardware. While most evolutionary robotics work is conducted in simulation due to the time and cost constraints of building physical robots, it is crucial to transition to hardware implementation to apply evolved robots to real-world problems.

The paper highlights the interplay between hardware constraints and the evolutionary process in designing an example system for evolved robot bodies. They found that every aspect of the hardware implementation introduces constraints that change the evolutionary space. Real-world limitations such as manufacturing and assembly constraints, as well as electrical power delivery, make it difficult to build certain robots or cause malfunctions in operation.

To address the challenge of hardware constraints, the researchers introduced methods of phenotype filtering and repair. However, this led to a degradation of robot population diversity and hindered the exploration space. The hardware constraints also limited the types of morphological variation that would be most useful in the target environment, reducing the ability of the evolutionary process to generate robots with effective adaptations.

The study concludes that designing a hardware platform for evolving robots requires a different mindset, where all design decisions should consider their impact on the viable phenotype space. It is also insufficient to solely evolve robots in simulation without detailed consideration of their hardware implementation, as the hardware constraints have a profound impact on the evolutionary space.

By understanding and addressing these challenges, researchers can enhance the applicability of evolved robots in real-world scenarios. The findings of this paper provide valuable insights for future developments in the field of evolutionary robotics.

Source: Angus M, Buchanan E, Le Goff LK, Hart E, Eiben AE, De Carlo M, Winfield AF, Hale MF, Woolley R, Timmis J and Tyrrell AM (2023) Practical hardware for evolvable robots. Front. Robot. AI 10:1206055. doi: 10.3389/frobt.2023.1206055