COMPUTATIONAL TASKS IN MEDICAL NANOROBOTICS - Bio-Inspired and Nanoscale Integrated Computing - page 413

legged ambulation, tank-tread rolling, amoeboid motion or inchworm
locomotion; control of histonatation (tissue swimming) and cytopenetra-
tion as required; cytocarriage (nanorobots controlling movement of
natural cells such as leukocytes from intracellular berths); and ex vivo
locomotion [1bj] including dental walking, epidermal locomotion, and
airborne nanoflight.
9. Computation. Control of all onboard computer systems [1d] including
CPUs, memory, and internal data transmission lines, and procedures for
switching between redundant computers or computer components; and the
management, synchronization and calibration of onboard clocks and
calendars [1bk]. Nanocomputers that employ standard dissipative architec-
tures (as opposed to energy-saving reversible computing architectures
[33-35]) will generate a high power density of waste heat, hence nanocom-
puters in larger nanorobots will commonly be throttled down from
maximum processing speeds during normal operations both to save energy
and to avoid excessive localized device heating. This is a principal design
limitation on medical nanorobot computers. Consequently, a good design
philosophy is to offload as many computational tasks as possible to in vivo
data processing devices located external to the nanorobot (e.g., fixed-
location tissue-embedded in vivo nanorobotic implanted nodes), or pre-
ferably to large ex vivo computers embedded in the physician's office,
the surgical suite, or the hospital infrastructure, with data transferred in
and out of the patient's body through fixed or mobile communications
networks [1bm].
10. Redundancy Management. Acceptable system reliability for populations
of trillions of cooperating medical nanorobots will require extensive
subsystem and component redundancy. Typically, tenfold redundancy
among mission-critical components appears sufficient to ensure accepta-
ble mission reliability [12g] in therapeutic applications. For example, a
good nanorobot design [4-8] may specify 10 times more sorting rotors,
sensors, or appendages for locomotion than are strictly necessary, with
the large surplus held in reserve as spares and backups. This implies that
another important computational function will be redundant systems
management [60, 61] and modeling [62]. Onboard computers must
continuously monitor the performance of all redundant components
and subsystems to determine whether or not any have failed, and if so,
to decide which backup system to swap in to replace the function
performed by the failed system. Studies are needed to define optimal
control protocols for redundant systems management. As a simple
example, in the case of the respirocytes which employ tenfold-redundant
onboard glucose-oxygen power stations, reliability simulations may be
useful to determine whether all ten power stations ideally should be run:
(a) at peak power on a rotating schedule, (b) at partial power on a
continuous basis, or (c) one at a time until failure, thereafter switching to