D(i) Backgthe Univeseveral yeconstrainetise in CMcircuits, wtions, dissand David(Figure 1) that consuretentive sexperienceperpetual lan all-digi60GHz anGSM broadisseminatwith Thom(ii) Visiovesting, segiven the including tdue to thelargely duetight coupultimately and integra1. Energ1mm3 nodand harvesand energy2. Node mode, monspent in stare typicalvolatile (Ngy of NV-the NV wr3. Wirelrequire timsources. Apendence desired frePERPDavid Wentz{kground and rsity of Miears of exped wireless seMOS imagerwireless circemination, ad Blaauw recomplete wumes 7.7µWleep state [1e in sub-threlifetime undital UWB trand a wake-upadcast signaltion of thesemas Schmid oon of the paensing, batterapid scalingthe aforemen challenges e to the powpling betweefactor into tation challengy resourcesdes is the missted on the ny harvesting standby pownitoring for andby, standlly power gNV) memorymemory is mrite process. less communme synchronAdditionally,on the technequency rangPETUAL Czloff, David BU{wentzlof; bexperience ochigan has erience deveensor nodesrs and solaruits, sensor and time synecently demwith solar harW in active 1]. This workeshold digitder moderateansmitter thp radio archis. Prabal Due motes into on time syncrticipants – ry, computing of ICs, anntioned funcof engineerwer requiremen underlyinthe minimumnges to realizs and powersmatch betwnode. Batteriis underpowwer and retea wakeup edby energy dated, howevy can be empmuch highernication – wnization betw antenna siznology. Antge. CUBIC-MMBlaauw, PraUniversity of laauw; prabaof the partica history oeloping tech. Our team hr arrays, enenode netwnchronizatiomonstrated a rvesting, batstate, and k has built oal circuits ae lighting coat consumesitecture to hutta co-develnearly 100 achronizationAs of todayng, and wirend proliferatictionality is ring the techents, energyng technologm size of the zing true cubr consumptiween the eneries in today’wered to achientive memoevent, rather dominates thver this causployed to savr than that ofwireless radioween two isze and efficiennas therefM WIRELEabal Dutta, TMichigan, Aal; thschmidcipants – Ourof collaborathnology for has combineergy-efficienorking and on. Dennis S9mm3 sensttery, and pr550pW in on nearly a dand memorieonditions. Das 12pJ/bit [2harvest a wirloped the Epapplicationsn and networky, there existeless commuion of MEMlimited to 1chnology smay storage, andgy, the peripsensor nodebic-mm sension – the prrgy consume’s systems aieve lifetimeory – sensor than in an ahe energy buses the contve state betwf RAM, and os typically csolated nodeiency are difore pose loESS SENSThomas SchmAnn Arbord; dmcs}@umr team at tion and energy-ed exper-nt digital applica-Sylvester sor node rocessing a data-decade of es. These pavid Wentzl2], as well areless synchpic [3] open m. He has recrking of wirets no compleunication in MS sensors, cm3 and largall enough tod system intpherals, and e. We envisisor nodes: rimary factoed by node reare typically es of even a nodes spendactive modeudget. To mintents of volaween wakeupcomplex perconsume reles, both signictated by raower limits oOR NODEmid, Dennis mich.edu power levelsloff has receas integratedhronization pmote platforcently publiseless sensor nete sensor na cubic-mmintegrating ager form faco fit these fotegration chad the applicaon the followor impeding esources and>100× largefew days. d most of thee. Because onimize standatile memoryup intervals, ripheral circlatively highnificantly dradio frequenon the node Figure 1. 0.5nWtemp. sensor, pbattery develoES Sylvester s are sufficieently demond patch antenpulse from erm and fosteshed several nodes [4][5]node includinm form factora complete ctors. This isform factors,allenges. Theation, all of wing technothe realizatd the energyer than we reir time in sof the relativdby power, cy to be lostbut the writcuits are needh active powraining battency, with litgeometry gW 9mm3 node wprocessor, solar aoped at UofM [1ent for nstrated nnas at xisting red the papers ]. ng har-r. Even system s partly , but is ere is a which logical tion of y stored require, tandby ve time circuits t. Non-e ener-ded for er, and ery re-ttle de-given a with and ].4. Software – writing software for this new computing class presents a number of challenges. The two classes of SRAM (volatile and NV) requires data partitioning and low-level memory man-agement. The high time and energy cost of paging underscore the need to track dirty data to reduce unnecessary write-backs. Limited energy requires an energy-aware OS scheduler. High concurren-cy from communications, image processing, and memory management, coupled with a small memory renders thread-based programming infeasible and event-driven programming preferable. (iii) Evidence that pursuing this vision will lead to major advances in the field – The technology being developed by our team at Michigan seeks to push the frontiers and enable the applications envisioned possible when perpetual, cubic-mm, wireless sensor nodes transitioned from science fiction into reality. Our team is collaborating with biologists, medical doctors, and other engineering disciplines to define applications that will benefit from pervasive, cubic-mm computing. Specific applications include intra-ocular and intra-cranial pressure sensing, intrusion monitoring, and inves-tigating the biotic effects of climate change on targeted animals. As cubic-mm platforms are adopt-ed by the sensor network community, many new and exciting applications will no doubt be realized by researchers years ahead of market forces. To that end, we identify the following three application themes that we seek to sufficiently support to enable third party research. Sensory skins cover sur-faces with a dense deployment of small, stick-on nodes that monitor the properties of the manifold itself or its surroundings including: detection and tracking of movement [6], detecting corrosion across metal surfaces [7], or monitoring EEG signals [8]. Thinking and linking gives everyday static and mobile objects sensing, computing, communication, and tracking ability. For example, tiny tags can be stitched into clothing for in-home elder care [9], smart waybills can report on the tempera-ture fluctuations a shipping package experiences [10], active baggage tags can locate a bag in an airplane fuselage [11], asset tracking can become as commonplace as asset tagging [12]. Implanta-ble intelligence gives visibility and voice to deeply embedded physical