Fire Monitoring Fireman in wild fire • report temperature within 100m of the moving user every 2s; Sensor Netw ork Services • temperature data must be no more than 1s old for Mobile Users Chenyang Lu Department of Computer Science and Engineering 1 2 Navigation through Fire Cargo Tracking Ship Train Truck Shipping Yard Customer, Shipper, Customs 3 4 Services for Mobile Users Services for Mobile Users Mobicast : information dissemination to moving areas. Mobicast : information dissemination to moving areas. � � MobiQuery : spatiotemporal query for mobile users. MobiQuery : spatiotemporal query for mobile users. � � Roadmap Query : navigation in dynamic environments. Roadmap Query : navigation in dynamic environments. � � MLDS : mobile agent directory service for tracking applications. MADS : mobile agent directory service for tracking applications. � � 5 6 1
A Motivating Application Fireman in wild fire • report temperature within 100m of the moving user every 2s; • temperature data must be no more than 1s old MobiQuery 7 8 Key Challenges Spatiotemporal Constraints � Limited power, memory and bandwidth. � Low duty cycle for extending network lifetime. Fireman in wild fire • report temperature within 100m of the moving user every 2s ; � Sleep schedule consists of cycles of long sleep period followed • temperature data must be no more than 1s old by short active period. � Lifetime of 450 days requires < 1% duty cycle. (based on Mica2 motes � Spatial constraints [J.Polastre et. al. Hot Chips ‘04]) � Query area moves with the user � All and only the sensors in the current query area should respond to the query 0.15s 0.15s 0.15s 0.15s active � Temporal constraints 14.85s 14.85s 14.85s 14.85s asleep � Result must be delivered within the current period cycle 1 cycle 2 cycle 3 cycle 4 � Result should not be older than data validity interval � Meeting the constraints are critical to the fireman’s safety! 9 10 Design Goals of MobiQuery Prefetching � Predict future query areas � Allows a mobile user to periodically collect sensor � Prefetching data from surrounding areas � Forewarn nodes ahead of time so that they wake up at � Meet spatiotemporal constraints the right time to deliver the sensor data � Deal with severe resource constraints � Query dissemination and data collection � Reduce storage cost � Nodes wake up in time and upload fresh data to user � Reduce communication overhead � Robust against unpredictable user movement 11 12 2
Wakeup Process Prefetching � Assuming backbone based power management (ex. � Greedy Prefetching CCP, GAF, Span) Sleeping node � Forewarn nodes in future query areas ASAP Wake-up Active node � Many routing trees set up simultaneously at T2 Forewarned node Wake-up � Prediction of far away query areas is likely to be wrong! at T1 � Just-in-time (JIT) Prefetching Wake-up at T3 � Forewarn nodes in future query areas at the right time 2 � Store and forward strategy 4 � Reduce network contention & storage cost Wake- 1 up at T4 � More robust to user motion changes 3 13 14 Directional Tree Creation (DTC) MobiQuery Protocols � Wake-up sensors ahead of the user � Directional Tree Creation (DTC) [ICDCS’05] � Create a new query tree in each query area � High overhead and network contention at high query rates � Aggregate sensor data and deliver to the user when the user � Requires knowledge of user motion profile reaches a query area Sleeping node Active node � Directional Tree Maintenance (DTM) [IPSN’05] Forewarned node � Reduces overhead and contention � Requires knowledge of user motion profile � Omni-directional Tree Creation (OTC) [IPSN’05] � Does not require knowledge of user motion profile 15 16 Generation of Motion Profile Directional Tree Maintenance (DTM) � Maintains a single moving tree rooted at the user � Local reconfiguration based on geographic location and user � Motion prediction motion profile Sleeping node Active node � Predict future user path based on movement history Forewarned node � Motion profile available after actual movement Query � Motion planning � Motion planner plans user path � Motion profile available before actual movement 17 18 3
Directional Tree Maintenance (DTM) Omni-directional Tree Creation (OTC) � Maintains a single moving tree rooted at the user � Local reconfiguration based on geographic location and user � No motion profile required motion profile Sleeping node Active node � Assume knowledge of maximum user speed Forewarned node 2 � Wake up sensors in a circular wake-up area 4 � Encloses sensors in all possible query areas in the 5 next few query periods � Create a tree when the user reaches a query area 1 3 19 20 OTC - Algorithm Analysis Sleeping node Active node � Derive parameters such that sensors in future query areas are woken up � early enough to respond to the query in time Query Query � late enough to reduce storage cost � DTM / DTC Query � Time to forward wake-up/prefetch message � Based on user velocity and sensor sleep period � OTC � Size of wake-up area � Based on knowledge of maximum user velocity and sensor sleep period 21 22 Simulations - Metrics Simulations - Settings � Run on ns-2 � Success ratio – fraction of queries � Backbone-based Power management: Coverage � that meet deadline Configuration Protocol � for which, the faction of sensors that respond in a � 200 sensors in a 450mx450m area query area is above a threshold � Query radius = 150m � Communication cost – total number of messages � Sensor nodes normalized by total number of query results � Communication Range – 105m � Sensing Range – 50m � Bandwidth – 2Mbps 23 24 4
Impact of Sleep Period Impact of User Speed 1 1 0.9 0.9 0.8 0.8 Success Ratio 0.7 Success Ratio 0.7 0.6 0.6 0.5 0.5 0.4 0.4 DTM 0.3 DTM 0.3 OTC 0.2 DTC OTC 0.2 No Prefetching 0.1 DTC 0.1 0 0 0 3 6 9 12 15 3~5 ms/s 6 ~ 10 m/s 16 ~ 20 m/s 36~40 m/s Sleep Period (s) User Speed (m/s) Figure 1. Effect on Success Ratio (T p =0.5s); threshold = 90% Figure 2. Effect on Success Ratio (T p =1s, T s =15s); threshold = 90% � No prefetching performs extremely poorly DTM and OTC successfully adapt to different speed ranges. � DTM > OTC >> DTC at long sleep periods 25 26 Impact of Location Error Impact of Motion Changes 1 1 0.9 0.9 0.8 0.8 Success Ratio Success Ratio 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 DTM Ts = 6s DTM Ts = 6s 0.3 0.3 DTM Ts = 12s DTM Ts = 12s 0.2 0.2 OTC Ts = 6s OTC Ts = 6s 0.1 0.1 OTC Ts = 12s OTC Ts = 12s 0 0 4 6 8 10 12 0 5 10 15 20 Number of Motion Changes Location Error (m) Figure 3. Effect on Success Ratio ; threshold = 80% Figure 4. Effect on Success Ratio (location error = 10m); threshold = 80% � OTC is not affected by inaccuracy in the motion profile; � OTC is not affected by user motion changes � DTM delivers 80% of the query results for location error of 15m. � DTM maintains a success ratio greater than 80% 27 28 Communication Cost Implementation on Motes 120 � Implemented DTC and DTM on 6x3 grid of Mica2 motes Tp=1s t R esul 100 Tp=0. 5s � Acroname PPRK robot carrying stargate was used to Q uery emulate the user 80 per 60 O verhead 40 20 0 O T C D T C D T M Figure 5. Under accurate motion profile (T s =9s) DTM << DTC < OTC 29 30 5
Summary � Meet stringent spatiotemporal constraints. � DTC - good performance for query periods ≥ 2s � DTM - lowest communication overhead. � OTC - most robust to location error and motion changes. � DTM and OTC successfully deliver >80% of results when Roadmap Query � query period = 1s for Navigation � <1% duty cycle � user changes direction every minute � location error = 15m 31 32 Problem Addressed Example Scenario T > Δ T � Mobile entity navigation in dynamic environments � Example dynamic environment � fire � Applications - search and rescue, evacuation B � Find a safe path from a start point S to a goal point G, G that is clear of fire. � Safe path - path where the maximum temperature along the path is below Δ T. C T CG < Δ T T AB < Δ T T AB > Δ T A S 33 34 Application Challenges New Challenges Heavy communication workload � Dynamic environments may change quickly. � Up-to-date information about the environment is Congestion, communication delay and data loss. required for successful navigation. � On-board sensors insufficient due to limited sensing Poor safety and navigation performance range. Solution? Efficient query that can provide required information Obtain up-to-date information about the at minimum communication cost surrounding area through a sensor network! 35 36 6
Recommend
More recommend
