Design Lessons from an Unattended Ground Sensor System

1
Design Lessons from an Unattended Ground Sensor
System

• Lewis Girod
• CS 294-1
• 23 Sept 2003

Center for Embedded Networked Sensing,
girod_at_cs.ucla.edu Work done at Sensoria Corp,
supported by DARPA/ATO contract
DAAE30-00-C-1055. Sensoria team R. Costanza,
J. Elson, L. Girod, W. Kaiser, D. McIntire, W.
Merrill, F. Newberg, G. Rava, B. Schiffer, K.
Sohrabi
2
Introduction
SPEC (J. Hill) 4MHz/8bit, 3K/0K

• Networked embedded systems come in a variety of
  sizes
• RAM is a primary tradeoff
• Costs power, cost to shutdown
• Enables greater complexity
• Enables development while postponing optimization

Mica2 (Berkeley/Xbow) 8MHz/8bit, 4K/128K
MK2 (UCLA/NESL) 40MHz/16bit, 136K/1M
SHM (Sensoria) 300MIPSFP/32bit, 64M/32M
3
Motivation for EmStar

• EmStar is a run time environment designed for
  Linux-based distributed embedded systems
• Useful facilities (process health/respawn,
  logging, emulation)
• Common APIs (neighbor discovery, link interface,
  etc)
• Designed for larger memory footprint (avoids hard
  limits)
• Many of the ideas and motivations for EmStar
  derived from our experience with SHM
• Modularity, robustness to module failure
• System transparency at low cost to developers
• Some parts of EmStar are used in SHM elsewhere
• Time Synch service
• Audio Server

4
System Objectives and Design
5
Objectives

• Unattended Ground Sensor (UGS) System
• Fully autonomous operation
• Ad-hoc deployment
• Scaling unit 50 nodes
• All operations and protocols local to 50 node
  region
• No global operations or context required

Fancy graphics taken from the official SHM
website
6
Adaptive and Self-Configuring

• Self-localizing without GPS acoustic ranging
• Build map of relative locations
• Adaptive/Resilient to environmental conditions
• e.g. wind, sunny days, background noise
• Self-assembled data network
• TDMA MAC layer
• Typically 10-hop diameter with 100 nodes
• Adaptive/Resilient to RF environment

In this application, GPS is avoided for security
reasons. In other applications, obstructions
and foliage can be an issue
7
Maintain coverage via Actuation

• Vigilant units detect failed unit(s)
• Remaining units autonomously move in to maintain
  coverage

8
Demonstration Requirements

• 200x50m outdoor field
• 100 nodes, 10m spacing
• Sunny afternoon
• 85F, 20 MPH wind
• No preconfigured state
• GPS-free relative geolocation to 0.25m
• Detect downed nodes, move to maintain coverage
  within 1 min

9
SHM Project Design Choices Optimize for rapid
development

• Concurrent HW/SW development
• Compressed schedule
• Aggressive scaling milestones
• Logistical problems with debugging system of 100
  nodes in 200x50m field
• Complex software required
• 150K lines C code
• 30 processes
• 100 IPC channels
• Power is not the driving constraint
• Continuous vigilance, rapid response are project
  requirements
• System lifetime target lt 1 day

10
System Configuration

• 300 MIPS RISC processor with FPU
• 64M RAM / 32M Flash
• 2 50kbps 2.4GHz data radios, TDMA,
  frequency-hopping, star-topology MAC, 63 hopping
  patterns
• 4 channels full-duplex audio
• 3-axis magnetometer / accelerometer
• 2 mobility units, with integrated thrusters
• Linux 2.4 kernel
• Optional wired ethernet (for Devel/Debug only)

11
Results Acoustic Ranging

• Ground truth was hand-surveyed, /- 0.5m
• Ranges not temperature compensated in demo
• Ranges with angles are more accurate
• Angle from TDOA of two or more ranges, must be
  consistent
• Bug discovered after the fact, caused large errors

12
Results Radio Utilization

• Graph shows traffic at three bases over a
  complete run
• Initial spikes
• Tree formation
• Lots of ranging
• Quiescent rates
• Heartbeats to detect down nodes
• Maintenance of trees and location
• Reaction to dynamics

13
Challenges in Implementation

• Dealing with a dynamic environment
• Adapt to wind, weather, RF connectivity
• Dealing with noise
• Rejecting outliers from timesync, ranges, angles
• Filtering neighbor connectivity, insignificant
  changes to range/angle
• Dealing with failure
• Node failure
• Infrequent crashes (e.g. FP exceptions from
  transient bad data)
• Fault tolerance at process boundaries, avoid
  ripple effect
• Dealing with complexity
• Cross layer integration vs. modularity.. Or both?
• What are the right set of primitives for
  coordination?

14
Case Study Acoustic Ranging
15
Basic TOF Ranging

• Basic idea
• Sender emits a characteristic acoustic signal
• Receiver correlates received time series with
  time-offsets of reference signal to find peak
  offset

16
Basic AOA Estimation

• 16 possible paths
• First pick best speaker
• Then estimate angle from TDOA of one or more
  consistent ranges

17
Acoustic Ranging, Version 1

• First cut implemented explicit cluster
  coordination protocol
• Lots of error cases to handle, hard to handle all
  efficiently
• Very timing sensitive (sync)
• Did not scale past 20 nodes
• Cant range across clusters
• Best acoustic neighbors may be in other clusters
• MLat merging algorithm is error prone
• Overuse of flooding
• Soft state reflood of cluster MLats and
  orientation data

18
V2 Decomposing AR
AR
19
Audio Sample Server

• Continuously samples audio, integrates to
  Timesync
• Eliminates error-prone Synchronized start
• Enables acquisition of overlapped sample sets
• Buffers past N seconds, exposes buffered
  interface
• Data access can be triggered after the fact
  relaxes timing constraints on trigger message
• Can process overlapping chirps by requesting
  overlapping retrievals, rather than having to
  pick one and ignore other
• Enables access so audio device from multiple apps
• Ranging can coexist with acoustic comm subsystem

Acoustic comm was developed as a backup channel
to be used in event of RF jamming
20
Inter-node Timesync RBS

• Key idea
• Receiver latency more deterministic than sender
• Thus, common receivers of a sender can be synched
  by correlating the reception times of senders
  broadcasts
• Its your only option if you dont control the
  MAC

For sender sync, senders must be in some other
senders broadcast domain
21
TimeSync Service

• Inter-node Sync Implementation of RBS
• Computes conversion params among all nodes in
  each cluster
• CH does computation, reports parameters to CMs
• Intra-node Sync Codec sample clocks
• Clock pairs reported by audio server
• Map time of DMA interrupt to sample number
• Outlier rejection and linear fit to find offset
  and skew estimate
• Yields more consistent result than synch start
• Multihop Time Conversion
• Graph of sync relations through system
• Conversion from one element to another requires
  path through graph. Gaussian error at each step
  sqrt(hops)

22
Hop-by-hop Time Conversion

• Problem
• Nodes have ability to convert within cluster but
  not outside
• Could continually broadcast conversion
  parameters BUT
• They are continuously varying
• Large amount of data to transmit across network
• Solution Integrate time conversion with routing
• Routing layer knows about packets that contain
  timestamps
• Convert timestamps en route
• At cluster boundaries
• At destination node
• Integrated with flooding
• Can fail if sync graph ? route

Unclear what the right API is here we simply
added code to flooding.
23
Reliable State Synchronization

• Problem
• Need to reliably broadcast the latest range data
  to N-hop away nodes, so they can build a
  consistent coordinate system
• Should have reasonable latency and low overhead
• V1 addressed this problem with periodic refresh
• Cluster heads retransmit Mlat tables every 15
  seconds
• Problems Traffic load from redundant sends,
  latency on msg loss
• Traffic load forced new protocol
• Send a hash when there was no change since last
  refresh
• If the hash has not been seen, request full
  version
• But, still has 15 second latency on lost data
• V2 introduced a Reliable State Sync protocol
  (RSS)

24
RSS Design

• Semantics
• Reliably converges on latest published state
• Does not guarantee client sees every transition
• Robust and Efficient, structurally similar to
  SRM/wb
• Based on reliable transfer of a sequenced log of
  diffs.
• Pruning of the log is done with awareness of log
  semantics (replaced or deleted keys are pruned)
• Per-source forwarding trees (MST of connectivity
  graph)
• Local repair, up to complete history, from
  upstream neighbor
• New or restarted nodes will download all active
  flows from upstream neighbor

25
RSS API

• Node X publishes current state as Key-Value Pairs
• Diffs are reliably broadcast N hops away from X
• Each node within N hops of X eventually sees the
  data X published
• API presents each nodes KVPs in its own
  namespace
• Caveat transmission latency, loss, edge of
  hopcount can cause transient inconsistencies

State Sync Bus
1
3
2
4
1 A1 1 B2 2 A3 2 C4
1 A1 1 B2 2 A3 2 C4
1 A1 1 B2 2 A3 2 C4
2 A3 2 C4
Note 2-hop publish from 1 doesnt reach 4
26
Putting it back together AR V2
27
AR V1 Event Diagram

• But, there are many error cases
• REQ lost?
• ACK lost?
• Bcast lost to some receivers?
• Bcast delayed in queue?
• Bcast lost to sender?
• CMs join two clusters may be busy ranging in
  other cluster.
• Inaccurate codec sync start?
• Interference in acoustic channel?
• Reply from sender lost?
• Reply from receiver(s) lost?
• How long to wait for stragglers?
• CH failure loses all ranges for cluster

Cluster Head (Coordinator)
Cluster Member (Sender)
Cluster Member (Receiver)
mlatd (CH)
AR (CH)
AR (CM)
AR (CM)
Reliability challenges The sender is the
linchpin an error in sender sync affects all
ranges to receivers, and replies from receivers
cant be interpreted without the sender reply.
If connectivity to sender is bad and the
broadcast is lost, all receivers waste CPU on a
useless correlation. Implementing reliable
reporting is made more complex because retxd
receiver replies must be matched to a past sender.
If not enough data for cluster mlat, request
ranging to specific missing cluster members
Send Range REQ to first CM in round robin order,
check busy
ACK with preferred start time
Bcast Range Start, specify code
Timestamp msg arrival. Sender delays before
starting to ensure rough sync, and reports exact
time offset from bcast to codec start
Acoustic Signal
Run correlation, report time offset from bcast to
detection in data

• Big complexity increase to
• Range across clusters
• Coordinate adjacent clusters
• Do regional mlats
• Average multiple sync bcasts

CH waits for stragglers
Report new ranges and notify mlat when round
robin thru CMs completes
28
AR V2 Event Diagram

• What can go wrong here?
• Collisions in acoustic channel.
• Flooded message delayed beyond audio buffering
  (16 seconds).
• Flooded message dropped for lack of sync
  relations along route.
• Node restart causes ranges to/from that node to
  be dropped.

Continuous Sampling
Continuous Sync Maintenance
Waiting for chirp notification
Waiting for chirp request
Waiting for enough data to compute mlat
mlatd
ar_send
ar_recv
syncd
audiod
If not enough data for mlat, request chirp and
wait for a while

• Key design points
• Encapsulate timing critical parts, no timing
  constraints on reliability.
• If a receiver cant sync to sender it wont
  attempt correlation.

Chirp audio (audiod on remote node records it)
Acoustic Signal
Flood Chirp notification message, with hop-by-hop
conversion at flood layer
Retrieve samples from buffer and correlate in
separate thread
Publish new range to N hop away neighbors
Try mlat again with new data in separate thread
29
Key Observations

• No coordination required
• Simplifying transport abstractions
• Continuous operation and service model

30
Key Observations

• No coordination required
• If mlatd doesnt have enough data it triggers
  chirping to start generating more data
• Exponential backoff on chirping with reset when
  data is lost.
• Simplicity of system lets designer focus on these
  details
• ar_send ar_recv are slaves to request and
  notify messages.
• Transparently, ar_recv can receive overlapping
  triggers and buffer the data for correlation
• Priority scheme decides the best order to process
  queued correlations, based on past
  success/failure and RF hopcount
• Simplifying transport abstractions
• Continuous operation and service model

31
Key Observations

• No coordination required
• Simplifying transport abstractions
• Flooding takes care of delivering a local time
• State Sync provides consistency for data input to
  mlatd
• Efficiently supports a potentially large number
  of keys (1000), enabling full regional mlat at
  each node (no merging)
• Mlat takes 10-15min, sync is consistent on that
  timescale
• Failure of one node only loses range data for
  that node
• Continuous operation and service model

32
Key Observations

• No coordination required
• Simplifying transport abstractions
• Continuous operation and service model
• Eliminates many inconsistencies and corner cases
• Reduces the number of states or modes
• Simplifies interfaces to services
• Recovery from faults without coordination just
  wait for stuff to start working again
• Service model supports multiple apps concurrently

33
The Catch

• Of course, the catch is power consumption
• Continuous operation can be wasteful
• Modularity can be less efficient than cross-layer
  integration
• Interesting questions
• How much is gained by fine-grained shutdown, plus
  the added coordination overhead, relative to more
  coarse grained shutdown and periods of continuous
  operation?
• For instance, the AR system could shut down after
  generating an initial map, and only wake up when
  something moves.

34
The End!
For more information on EmStar,
see http//cvs.cens.ucla.edu/emstar/
35
Design Evolution

• Initial design strategy shortest path first
• Modular decomposition according to best guess at
  time
• Making a full-blown, generalized service is much
  more work than a one-off feature so tradeoff
  considered case by case
• Problem As more is learned these tradeoffs fit
  more poorly
• Unmanageable complexity to address problems
• Redesign
• Factor out common components
• Plan for known scaling problems
• Remaining modules are of manageable complexity,
  yet usually achieve a more complete and correct
  implementation
• More sophisticated inter-module dependencies

36
Radios

• Each node has two radios
• TDMA, frequency hopping radios
• 63 hopping patterns
• Each radio can lock to one pattern
• Patterns are independent channels
• Bases on same pattern tend to be desynchronized
• Base/Remote (star) topology
• Base synchronizes TDMA cycle, remotes join

37
TDMA Slot Scheme

• Each frame contains 1 transmit slot for the base
  and 1 transmit slot for each remote
• Slot size implies MTU
• Frame size is a constant
• Base slot size is fixed 70 byte MTU
• Number of remotes inv. prop. to remote MTU
• Practical MTU (40 bytes) ? 8 node clusters

14ms
Base
S
Base Slot
Remote 1
Remote 2
Remote 3
38
Packet Transfer

• Broadcast capability
• Base can use its slot to send a broadcast to all
  remotes, or a unicast to a single remote
• Remotes can send only unicasts to base
• Link layer retransmission
• MAC implements link layer ACKs for unicast
  messages, and configurable retransmission

39
Breach Healing
In this application, healing is intended only
to address breaches created by dying nodes, not
preexisting breaches. Other algorithms might also
be useful, e.g. density maintenance, but were not
implemented here.