This manuscript is now published!
Figure: (A) Animals outperform robots in range, agility, and robustness. (B and C) The power, frame, actuation, sensing, and control of machines often outperform robots when those components are considered separately.
We compare measured performance values from the literature on animal physiology and robot design. Since many properties vary with size and shape, we consider representative examples on the scales of cockroaches, cats, and humans. We investigated whether differences in five subsystems of running animals and machines explained the different range, agility, and robustness capabilities of animals versus machines. These subsystems include power, frame, actuation, sensing, and control. Of particular relevance to the MURI are the sensing and control subsystems. In terms of sensing capabilities, biological and engineered photoreceptors are comparable in their overall counts and ability to detect visual stimuli. While engineered mechanoreceptors can detect much smaller stimuli than biological ones, biology’s ability to integrate staggering numbers of mechanosensors distributed throughout bodies, including the electrical system needed to innervate the sensors, is remarkable. In terms of control capabilities, computer networks vastly exceed the performance of nervous systems in latency and bandwidth of communication and computation, but artificial neural networks are at a significant disadvantage relative to the size and connectivity of biological networks. Animals cannot practically decrease sensorimotor delay by the orders of magnitude that would be required to compete with robots’ communication channels; this fundamental limit surely affects control strategies, for instance by favoring the use of internal models. Although neuromorphic circuits will continue to increase in complexity, it remains to be seen whether bigger brains are better and how to make the most effective use of the limited brainpower available to robots in the meantime.
Overall, we conclude that not one subsystem significantly limits the general performance of autonomous robots compared to their biological counterparts (Figure). Instead, it is the integration of these components that seems to be limiting. This suggests a new approach to integrated design is needed that emphasizes targeting emergent properties of the composite systems rather than breaking the problem down into subsystems. We suggest a few possible alternative strategies, such as integrated control policy design, to achieve these ends.
