In previous articles, I explained how bioengineer Stuart Burgess’s research demonstrates the exquisite design behind vertebrate limbs, and how it overturns the claim that some limbs appear poorly designed (here, here, here). Here, I will describe the genius behind the limbs’ control systems, which are even inspiring innovation in the control systems of robotic manipulators. In addition, the stringent requirements of motor control systems further overturn evolutionary narratives and bolster the case for design. 

Control System Challenges

Designing control systems for complex robotic limbs has proven extremely difficult. In “Intelligent control of robotic manipulators: a comprehensive review,” Rawat et al. (2022) explain that the challenge is due to a control system requiring a specific layout with fine-tuned parameters that perfectly matches the architecture and dynamics of its robotic manipulator. A perfect matching is needed since the precise control of motion is essential to perform complex tasks.  

Engineers must employ highly sophisticated methods to direct robotic machinery, and each approach has faced limitations. Conventional approaches attain stability, accuracy, and flexibility, but they cannot accommodate many degrees of freedom (e.g., large number of joints and digits) as seen in biological limbs. Conversely, learning-based approaches can manage many degrees of freedom, but they lack stability and flexibility. Alepuz et al. (2024) summarize the difficulty in their article “Brain-inspired biomimetic robot control: a review”:

Conventional model-based control approaches guarantee strong stability properties of the controlled system and prescribed accuracy, even in the presence of structured and unstructured uncertainties. However, their design complexity scales very poorly with dimensionality [number of degrees of freedom] and, therefore are difficult to generalize, maintain and tune in connection to complex robot tasks. On the other hand, relying on model-free or learning-based solutions, such as machine learning and statistical modeling methods can efficiently manage extensive system dimensions. Yet, they come with heavy computationally demands, struggle with adaptability to different scenarios and lack assurance in stability and robustness.

Brain-Inspired Robotic Control

Robotics engineers have looked to the human nervous system to learn how it precisely maneuvers limbs with many degrees of freedom to perform complex tasks in uncertain environments. Alepuz et al. (2024) describe the nervous system’s architecture as follows:

The brain’s motor system comprises specialized areas dedicated to distinct functions in controlling movement. These regions follow a hierarchical arrangement: higher-level domains oversee broader tasks with considerable abstraction, while lower-level segments focus on individual muscles, delivering precise signals tailored to the task’s specifics. Complementing this structure are additional side structures (side loops) responsible for regulating signals within the descending pathways of this hierarchical system. 

Brain-inspired control systems employ neural networks with millions of neurons that mimic the human nervous system’s design. The neurons require highly specific interconnections with each other and with a multitude of sensors. The system will only function after it meets exacting requirements. 

Challenge to Evolution

The high minimal complexity of and tight constraints on vertebrate limbs’ sensory and control systems preclude any possibility of their having emerged through an undirected evolutionary search. The sensors themselves involve numerous intricate biological structures that measure such variables as motion, strain, and pressure (here, here). The severe challenges of navigating the evolutionary fitness landscape associated solely with the mechanics of vertebrate limbs are dramatically compounded when considering both limb mechanics and the control system. (For a summary of evolutionary fitness landscapes, see my article “Fitness Landscapes Demonstrate Perfection in Vertebrate Limbs Resulted from Intelligent Design.”) 

Cheney et al. (2016) detail how evolutionary searches that consider both stall on a less-than-ideal limb architecture (aka morphology) and then only adjust the control system for that suboptimal design. One limb design can never incrementally transform into a significantly different design without intelligent intervention. Nygaard et al. (2017) summarize the barrier as follows:

Optimizing robot morphologies together with control systems is a difficult task, and we have only so far seen relatively simple results, even though significant amounts of computational resources have been allocated. One of the reasons for this difficulty is the increased dimensionality of the search space that comes with the freedom to design the body, in addition to the control system. In effect this requires a much larger amount of exploration to find the desired quality of solutions. However, another aspect of the difficulty is that the co-evolution of morphology and control also leads to a much more difficult search — the search landscape is much more rugged, and small changes in the morphology can easily offset the performance of a previously found good body-controller combination.

Different vertebrate limbs represent fundamentally different design patterns that operate within mutually exclusive constraints. Even if one limb suddenly transformed into another (e.g., wolf forelimb into a whale flipper), the new limb would prove useless until its control system was entirely reengineered to match the new limb mechanics. Only an intelligent agent can simultaneously engineer a novel limb with a new control system that matches it.