The Cassini Resource Exchange
Client: Jet Propulsion Laboratory & NASA
On October 15, 1997, the National Aeronautics and Space Administration (NASA) launched the Cassini-Huygens spacecraft on a Titan IV-B/Centaur rocket for a seven year voyage to Saturn. The spacecraft and its subsystems carried an array of sophisticated instruments and cameras along with a probe (Huygens) that was sent to the surface of Saturn’s largest moon Titan. While the mission is a scientific treasure trove, it was also a management success never experienced in the history of planetary missions: the mission came in at cost and required fewer resources than planned. To understand the management process used on Cassini we first describe the history leading up to the Cassini mission and the need for a radical departure in its management.
In general, planetary missions conducted by NASA are managed by the Jet Propulsion Laboratory (JPL) in Pasadena, California. JPL proposes missions which NASA funds and it is the responsibility of JPL to carry out the missions. This involves selecting instruments that will fly on a spacecraft that is built by JPL and its contractors. Instruments are typically built by University teams or research centers. The instruments are highly sophisticated hardware and software systems and represent the career of the principal investigator (PI) who is responsible for the design, development and operation of the instrument.
Prior to Cassini, JPL was being chastised for its handling of the Mars Observer (MO) mission which had a cost overrun of 150%, was delayed by 4 years and each instrument had significant power and mass growth requirements that spurred the cost overruns and schedule delays. The question at hand was what could be the cause of such instrument growth and how could it be prevented in the future.
Unfortunately MO was not an aberration but standard fare for planetary missions. Because the instruments are typically state of the art builds, they are besieged with “technological” uncertainty. The management method of choice during this timeframe was called “management by margins.” It was postulated that due to the uncertainty during instrument development, an insurance pool of funds, mass and electric power should be held to assist the spacecraft and instrument that encounter problems during development. In practice this meant that some percentage (margin) of the mission resource envelope of budget, mass and electric power should be held by project management to solve the problems that will “naturally” arise. Since instruments are built by scientists who have an entire career riding on the quality of the data returned by their instrument they have a “natural” incentive to develop an instrument that has state of the art capabilities. The existence of an insurance pool is too much of a temptation to the instrument builder to push the technology envelope since any “bad luck ” will be covered by project management. This type of behavior is referred to in the economics literature as moral hazard. If this were truly insurance, some instruments would have good luck and some bad luck, but on average the insurance would cover the bad luck. However, in planetary missions everyone has universally “bad luck” in instrument development. It is often noted by instrument PIs that if you are going to have problems in development you should have them early while there is still margin to capture. On the other hand “good luck” usually means that your unused resources will be allocated to solve the problems of an instrument in trouble, or as once stated by an instrument PI, “no good deed goes unpunished “.
Given this incentive problem what is management to do? The margins are akin to the dilemma of the commons. The project could “monitor” instrument development more closely having parallel teams looking over shoulders. This is costly to say the least. Voluntary control seems not to work in this type of environment. Thus, the only remaining possibility is to privatize the margins and allow each instrument to determine its own fate and make resource tradeoffs. This is precisely what was done for Cassini. It was proposed that no margins be held and that individual resource envelopes would be assigned to each PI to be used to develop their instrument. Each PI was in charge of their own destiny and if they could not build an appropriate instrument for the resource envelope they were assigned, their instrument was subject to de-selection and eliminated from the payload. This was a very stark management process, one unfamiliar to the science community. However, an issue arose as to what PIs were able to do with their resources. What if new information arose that would allow them to build a lighter instrument (require less mass) but at the expense of a larger power requirement? It was proposed that the PIs should be allowed to trade resources among themselves. The next question that arose was what trading process should be instituted. Arlington Economic principal, David Porter, helped design and implement the trading system used on Cassini. However, all trades had to be swaps of resources. NASA would not allow external monetary transfers between participating organizations. Thus, a barter exchange was developed and tested.
To assist barter trading A Computer Assisted Resource Exchange (CARE) was created that allowed for the simultaneous execution of combinations of swaps without going through the full sequence of trades. Second, if a participant could not find a desirable trade with the bids in the system, CARE offered information about how to modify a participant’s bids to be able to complete a trade. Lastly, if a participant were interested in only trading in a particular subset of resources, CARE allowed the filtering of information to reduce the number of bids to examine. This was the first market ever created on the Internet.
The results for the mission instrument management were stellar. All the instruments were launched and the PIs managed their mass so well that the instruments were 7% under their mass allocation! We should point out that at the beginning of the process several PIs announced that this would never work because mass is always in short supply at the end of development and that it would never be traded or that its price would be astronomical. Figure 1 shows the trading activity in mass over the life of the market. It is clear that the predictions on mass were somewhat exaggerated.