Yesterday’s thoughts on self-repairing chips, as demonstrated by recent work at Caltech, inevitably called Project Daedalus to mind. The span between the creation of the Daedalus design in the 1970s and today covers the development of the personal computer and the emergence of global networking, so it’s understandable that the way we view autonomy has changed. Self-repair is also a reminder that a re-design like Project Icarus is a good way to move the ball forward. Imagine a series of design iterations each about 35 years apart, each upgrading the original with current technology, until a working craft is feasible.

My copy of the Project Daedalus Final Report is spread all over my desk this morning, the result of a marathon copying session at a nearby university library many years ago. These days you can skip the copy machine and buy directly from the British Interplanetary Society, where a new edition that includes a post-project review by Alan Bond and Tony Martin is available. The key paper on robotic repair is T. J. Grant’s “Project Daedalus: The Need for Onboard Repair.”

Staying Functional Until Mission’s End

Grant runs through the entire computer system including the idea of ‘wardens,’ conceived as a subsystem of the network that maintains the ship under a strategy of self-test and repair. You’ll recall that Daedalus, despite its size, was an unmanned mission, so all issues that arose during its fifty year journey would have to be handled by onboard systems. The wardens carried a variety of tools and manipulators, and it’s interesting to see that they were also designed to be an active part of the mission’s science, conducting experiments thousands of kilometers away from the vehicle, where contamination from the ship’s fusion drive would not be a factor.

Even so, I’d hate to chance one of the two Daedalus wardens in that role given their importance to the success of the mission. Each would weigh about five tonnes, with access to extensive repair facilities along with replacement and spare parts. Replacing parts, however, is not the best overall strategy, as it requires a huge increase in mass — up to 739 tonnes, in Grant’s calculations! So the Daedalus report settled on a strategy of repair instead of replacement wherever possible, with full onboard facilities to ensure that components could be recovered and returned to duty. Here again the need for autonomy is paramount.

In a second paper, “Project Daedalus: The Computers,” Grant outlines the wardens’ job:

…the wardens’ tasks would involve much adaptive learning throughout the complete mission. For example, the wardens may have to learn how to gain access to a component which has never failed before, they may have to diagnose a rare type of defect, or they may have to devise a new repair procedure to recover the defective component. Even when the failure mode of a particular, unreliable component is well known, any one specific failure may have special features or involve unusual complications; simple failures are rare.

Running through the options in the context of a ship-wide computing infrastructure, Grant recommends that the wardens be given full autonomy, although the main ship computer would still have the ability to override its actions if needed. The image is of mobile robotic repair units in constant motion, adjusting, tweaking and repairing failed parts as needed. Grant again:

…a development in Daedalus’s software may be best implemented in conjunction with a change in the starship’s hardware… In practice, the modification process will be recursive. For example the discovery of a crack in a structural member might be initially repaired by welding a strengthening plate over the weakened part. However, the plate might restrict clearance between the cracked members and other parts, so denying the wardens access to unreliable LRUs (Line Replacement Units) beyond the member. Daedalus’s computer system must be capable of assessing the likely consequences of its intended actions. It must be able to choose an alternative access path to the LRUs (requiring a suitable change in its software), or to choose an alternative method of repairing the crack, or some acceptable combination.


Image: Project Daedalus was the first detailed study of an interstellar probe, and the first serious attempt to study the vexing issue of onboard autonomy and repair. Credit: Adrian Mann.

The Probe Beyond Daedalus

Robert Freitas would follow up Daedalus with his own study of a probe called REPRO, a gigantic Daedalus capable of self-reproduction from resources it found in destination planetary systems. Another major difference between the two concepts was that REPRO was capable of deceleration, whereas Daedalus was a flyby probe. Freitas stretched the warden concept out into thirteen different species of robots who would serve as chemists, metallurgists, fabricators, assemblers, repairmen and miners. Each would have a role to play in the creation of a new probe as self-replication allowed our robotic emissaries to spread into the galaxy.

Freitas would later move past REPRO into the world of the tiny as he envisioned nanotechnology going to work on interstellar voyages, and indeed, the promise of nanotech to manipulate individual atoms and molecules could eventually be a game-changer when it comes to self-repair. After all, we’d like to move past the relatively inflexible design of the warden into a system that adapts to circumstances in ways closer to biological evolution. So-called ‘genetic algorithms’ that can test different solutions to a problem by generating variations in their own code and then running through generations of mutations are also steps in this direction.

One thing is for sure: We have to assume failures along the way as we journey to another star. Grant sets a goal of 99.99% of all components aboard Daedalus being able to survive to the end of the mission. This was basically the goal of the Apollo missions, though one of those missions suffered only two defects, equivalent to a 99.9999% component survival rate. Even so, given the need for repair facilities and wardens onboard to fix failing parts, Grants figures show that a mass of spare components amounting to some 20 tonnes needs to be factored into the design.

It will be fascinating to see how Project Icarus manages the repair question. After all, Daedalus was set up as an exercise to determine whether a star mission was feasible using current or foreseeable science and technology. With the rapid pace of digital change, how far can we see ahead? If we’re aiming at about 35 years, do we assume breakthroughs in nanotechnology and materials science that will make self-healing components a standard part of space missions? Couple them with advances in artificial intelligence and the successor to Daedalus would be smaller and far more nimble than the original, a worthwhile goal for today’s starship design.

The two papers by T.J. Grant are “Project Daedalus: The Need for Onboard Repair,” Project Daedalus Final Report (1978), pp. S172-S179; and “Project Daedalus: The Computers,” Project Daedalus Final Report (1978), pp. S130-S142.