Artificial intelligence is defined as the ability of machines to meaningfully adapt their behaviour to their surroundings. In the minds of the public, “artificial intelligence” focuses on human-like intelligence. However, many significantly simpler beings, such as insects, display intelligent behaviour. In the light of their reduced size and complexity, these animals in particular provide the ideal template for understanding the biological basis of behaviour and incorporating those findings into technological applications. Applications of this kind – i.e., technological devices, which consciously and autonomously interact with their environment – already play a vital role in contemporary space exploration. In the future, the complexity of missions will continue to increase, and with it the demands placed on artificially intelligence space probes, maintenance systems, and mobile reconnaissance units.
Mobile reconnaissance units such as the Mars Explorer are a textbook example of the application of artificial intelligence in space. In parallel with the scientific success of such robots in the cosmos, self-driving tractors or cars are being developed back on earth. Naturally, artificial intelligence plays a significant role beyond mobile reconnaissance units of this kind. Space probes are at least partially dependent on artificial intelligence when investigating such distant targets. Their missions are not just limited to navigation and the proper use of the various measuring instruments. The perfect functioning and interaction of the various technical components also requires continuous monitoring; artificial intelligence now takes some of the burden away from human ground controllers in relation to routine tasks such as these.
Self-driving units, such as the Mars Explorer, have achieved major successes with one important limitation: they are restricted to relatively smooth ground. The use of insect-like robots could be considered as an alternative. From a relatively large number of different projects, it is worth mentioning a robot from the University of Berkeley and another from the University of Bielefeld. The motion system of Hektor, the Bielefeld robot, is modelled on that of a rod grasshopper, while the Berkeley robot imitates a cockroach. Cockroaches can insert themselves into the narrowest cracks and the Berkeley robot shares this special ability. One particularly notable characteristic of the Bielefeld robot is the fact that its six different legs can be moved relatively independently of each other. Real stick insects also lack central control of their motion. Each leg overcomes obstacles that it encounters, and which stand in the insect’s way, in its own right. Meanwhile, the robot can find its way over rough surfaces just as effectively as a stick insect. It doesn’t even need a camera to provide an image of its surroundings – as such it can crawl blindly over the terrain.
In addition to the concrete example of motion in insects, significantly more generalised biological principles form the basis for technical solutions. Evolution involves the interplay of variation and selection that leads to the development of complex, highly successful organisms. This development occurred without any overarching regulation or planning. In this respect, it is similar to the phenomenon of the swarm, in which large groups of organisms (such as insects) achieve complex results that extend far beyond the abilities of the individual organisms. While the application of swarming rules to simply constructed robots is still in the development phase, evolutionary algorithms are already proving themselves in practice. In the search for an optimal pathway to solve a particular problem, solutions are randomly varied, with each one being tested for its usefulness. This approach is frequently used to address problems involving the interpretation of complex and large datasets, as in the neural networks described below.
Nerve cells, or neurons, are the biological building blocks of the nervous system in animals. In such systems, each individual neuron is connected to many others via complex pathways. The tree-like, branched dendrites receive signals from other cells. The more uniform axons are used to transmit the integrated signal that is received. The result is a complex network that processes the signals received from the sensory organs and triggers a reaction in the organism. Connections within the networks determine which stimulus patterns trigger which reactions. As these connections can be changed, e.g., due to repeated practice, neural networks are able to learn, and this is a highly sought-after characteristic of neural networks in the field of robotics. Simplified models of these networks (artificial neural networks) offer practical, relevant solutions in recognising and sorting complex patterns, such as human faces or handwriting. Artificial networks are similarly suited to the interpretation and validation of technical data or the search for a suitable landing zone by reviewing camera shots.
The example of the Mars Explorer shows that current space exploration is necessarily dependent on artificial intelligence. Technical progress in the field of artificial intelligence has enabled projects that would have been inconceivable just a short time ago. Organisations such as NASA or ESA have shown a corresponding degree of interest in developing artificial intelligence, as even a cursory glance at their research programmes will demonstrate. In this way, space exploration is contributing the development of self-driving vehicles on earth, while the development of self-driving vehicles is the driving force between research on the moon, Mars, and Venus.