(2.3.

5 Case Study: TinyOS and nesC

Event-Based Programming and Modularity:

● TinyOS supports modular, event-based programming, essential for sensor node concurrency
within resource constraints.
● Components encapsulate related functionality, like handling a radio interface or computing
routes, including state information, task code, and event/command handlers.
● Components are hierarchically arranged: low-level components interact with hardware, high-
level components form the application.

Component Interaction:

● Commands flow from high-level to low-level components; events flow upward.

● Both command and event handlers run to completion, performing simple tasks without
blocking.
● Tasks, triggered by handlers, run to completion but can be interrupted, eliminating the need
for stack management and ensuring atomicity.

Scheduling and Feedback:

● TinyOS uses a power-aware FIFO scheduler, shutting the node down when idle.
● Feedback is managed using split-phase programming: commands are followed by separate
completion events, enabling concurrency.

Interfaces in nesC:

● nesC allows defining interface types, grouping related commands and events, simplifying
split-phase programming.
● Components provide and use interfaces, facilitating modular design.
● Primitive components (modules) and configurations (wired components) ensure compatibility
through interface matching, checked by the compiler.

Example Components:

● Timer component: handles commands (init, start, stop) and events (fire, fired), interacting with
a hardware timer through setRate commands.
● CompleteTimer: integrates TimerComponent and HWClock, exposing StdCtrl and Timer
interfaces.

Practical Use and Extensions:

● TinyOS, with nesC, supports core OS functionalities, communication protocol stacks, and
application functions, proving effective in creating small, specialized components.
● It has become the standard platform for WSNs, available on various platforms.
● Extensions like virtual machines (for over-the-air reprogramming) and TinyDB (sensor
network as a relational database) expand its capabilities.
Example nesC Code:

Listing 2.1: Defining modules and interfaces

interface StdCtrl {
command result_t init();
}
interface Timer {
command result_t start(char type, uint32_t interval);
command result_t stop();
event result_t fired();
}
interface Clock {
 command result_t setRate(char interval, char scale);
event result_t fire();
}
module TimerComponent {
provides {
interface StdCtrl;
interface Timer;
}
uses interface Clock as Clk;
}

Listing 2.2: Wiring components to form a configuration

configuration CompleteTimer {
provides {
interface StdCtrl;
interface Timer;
}
implementation {
components TimerComponent, HWClock;
StdCtrl = TimerComponent.HWClock;
Timer = TimerComponent.Timer;
TimerComponent.Clk = HWClock.Clock;
}
}

TinyOS and nesC together provide a robust, modular, and efficient foundation for WSN development,
enabling scalable and maintainable sensor network applications.

2.3.6 Other examples

Apart from TinyOS, there are several other execution environments or operating systems for WSN
nodes:

Contiki:

● Contiki is an operating system for WSNs, known for its portability across various hardware
platforms.
● It implements a TCP/IP stack, even on platforms with severely restricted resources.

eCos:

● eCos (embedded Configurable operating system) is another operating system used in WSN
environments, offering configurable and modular capabilities.

Mantis:

● The Mantis project is another example of an operating system designed for WSN nodes,
focusing on ease of use and support for various hardware.
These alternatives provide various features and capabilities, each suited for different WSN
requirements and constraints. )

(5.1 Fundamentals of (wireless) MAC protocols

● Introduction and Background: MAC (Medium Access Control) protocols are fundamental in
wireless communication, managing access to shared communication channels. They have
been extensively researched for over 30 years, with literature covering general and wireless-
specific protocols.
● Energy Considerations: Earlier MAC protocol research prioritized factors like throughput
and delay over energy efficiency. However, the emergence of WSNs elevated energy
efficiency to a primary design concern, given the constrained power resources of sensor
nodes.
● Inheritance in WSNs: MAC protocols used in WSNs inherit challenges and solutions from
broader wireless networks. They adapt strategies to manage energy consumption, optimize
channel access, and ensure reliable communication within the context of sensor node
constraints.

MAC protocols continue to evolve, integrating energy-efficient strategies to enhance network

performance and longevity in wireless sensor networks.

5.1.1 Requirements and design constraints for wireless MAC protocols

1. Performance Requirements:
○ Throughput efficiency: Maximize data transfer rate.
○ Stability: Ensure consistent performance.
○ Fairness: Equal access for all nodes.
○ Low access delay: Minimize time between packet arrival and transmission attempt.
○ Low transmission delay: Minimize time from packet arrival to successful delivery.
○ Low overhead: Minimize MAC headers, trailers, collisions, and control packet
exchanges.
2. Collision Issues:
○ Collisions occur when multiple nodes send packets simultaneously, causing packet
decoding errors and retransmissions.
3. Time-Critical Applications:
○ Require deterministic or stochastic guarantees on delivery time or minimum data
rates.
○ May prioritize important packets using priorities.
4. Wireless Medium Challenges:
○ Issues like time-variable, error rates due to slow/fast fading, path loss, attenuation,
and manmade/thermal noise affect performance.
○ Bit error rates can range from 10-3 to 10-2 depending on factors like modulation
schemes, frequencies, and tx-rx distance.
5. Path Loss and Signal Strength:
○ Received power decreases with distance b/w tx & rx node, limiting communication
range.
○ Hidden-terminal and exposed-terminal problems can arise in wireless networks when
2 nodes are out of reach.
6. Hidden-Terminal Problem:
○ Occurs in CSMA protocols when a node cannot detect another’s transmission,
causing needless collisions.
○ A & C are out of range. A starts txng packet to B. C senses medium, finds it idle since
it cant hear A. C txs packet to B. Signal collision occurs at B. Both packets are
useless.
 7. Exposed-Terminal Problem:
○ Happens when a node unnecessarily defers transmission due to another node's
transmission, wasting bandwidth.
○ B txs to A. C txs to D. But C suppresses its tx because C can hear B and thinks
medium is busy.
8. Solutions to Terminal Problems:
○ Busy-tone solutions and RTS/CTS handshake (IEEE 802.11 standard and
MACA/MACAW protocols).
9. Collision Detection in Wired vs. Wireless:
○ Wired: Collision detection(tx detects collision at rx and aborts packet txn) is feasible due to
low attenuation and similar SNRs at transmitter and receiver. Eg: Ethernet’s CSMA/CD
protocol to ↑ throughput efficiency.
○ Wireless: Collision detection is difficult due to half-duplex mode and varying
interference conditions.
10. Spectrum Sharing and Coexistence:
○ WSNs often share ISM bands (e.g., 2.4 GHz) with other systems like IEEE 802.11,
Bluetooth, and IEEE 802.15.4, leading to coexistence challenges.
11. Traffic Load Patterns:
○ Continuous monitoring applications expect low, periodic traffic (Eg: time dependent
temperature distribution in forests).
○ Event-driven applications face idle periods followed by bursts of data needing high
MAC efficiency (e.g., wildfire observation).

5.1.2 Important classes of MAC protocols

MAC (Medium Access Control) protocols for wireless networks can be categorized into three main
classes: fixed assignment protocols, demand assignment protocols, and random access protocols.

Fixed Assignment Protocols:

● Description: These protocols allocate resources (such as time, frequency, codes, or space)
to nodes for exclusive use over a long-term basis, often minutes or hours.
● Scalability:
○ Changes in Topology:
■ Nodes dying or new nodes being deployed
■ Mobility of nodes
■ Changes in load patterns
○ Signaling Mechanisms:
■ Needed to renegotiate resource assignments in response to topology
changes
○ Scalability Concerns:
■ Effectiveness of protocols in adapting to frequent and dynamic changes.
● Examples:
 ○ Time Division Multiple Access (TDMA):
■ Divides time axis into fixed-length superframes which is divided into fixed
number of time slots, with each node allocated specific slots for transmission.
■ Requires tight time synchronization to avoid overlapping of signals in
adjacent time slots.
○ Frequency Division Multiple Access (FDMA):
■ Divides the frequency spectrum into subchannels, each assigned to a node.
■ Requires frequency synchronization, NB filters, rx ability to tune to channel
used by tx.
■ Tx is more complex.
○ Code Division Multiple Access (CDMA):
■ Nodes spread their signals over a wide bandwidth using unique codes to
distinguish transmissions.
■ Rx needs to know code used by Tx.
■ All other parallel txns using other codes appear as noise.
■ Code management is crucial.
○ Space Division Multiple Access (SDMA):
■ Utilizes spatial separation of nodes, requiring antenna arrays and
sophisticated signal processing.

Demand Assignment Protocols:

● Description: Resources are allocated on a short-term basis, typically for the duration of a
data burst. These protocols can be centralized or distributed.

Centralized Demand Assignment Protocols

● Description:
○ Resources allocated on short-term basis (duration of data burst)
○ Central node manages resource allocation based on node requests
○ Examples: HIPERLAN/2, DQRUMA, MASCARA protocols
○ Polling schemes are also included in this category
● Operation:
○ Nodes send bandwidth requests to central node using contention-based
random access protocols on a dedicated signaling channel. Central station
can poll its associated nodes for bandwidth allocation requests.
○ Central node accepts/rejects requests and confirms allocation
○ Allocated resources include time slots (in TDMA) or duration of allocation.
● Resource Management:
○ Central node remains powered on constantly
○ Handles resource allocation and deallocation implicitly
○ Nodes piggyback requests on data packets to avoid separate signaling
● Energy Considerations:
○ Central node consumes significant energy due to constant operation
○ Suitable when energy-unconstrained nodes are available
● Examples:
○ IEEE 802.15.4 protocol (Section 5.5)
○ LEACH protocol (Section 5.4.1) rotates central station duties among nodes

Distributed Demand Assignment Protocols

● Description:
○ Example: Token-passing protocols like IEEE 802.4 Token Bus
○ Nodes gain transmission rights via token frame reception
○ Token frame is rotated among nodes organized in logical ring structure on
broadcast medium
● Challenges:
○ Maintaining logical ring in error-prone or wireless media
○ Significant signaling required for ring management and error correction
 ● Operation:
○ Nodes must receive token to initiate transmissions
○ Token circulation time varies, requiring nodes to remain active
○ Management procedures handle node inclusion/exclusion and token loss
● Limitations:
○ Vulnerable to channel errors and frequent topology changes
○ High signaling overhead due to token management and maintenance

Random Access Protocols: (Contention based protocols)

● Description: Nodes operate independently with no coordination, using random
elements(random packet arrival times and random timer values) to access the medium.
● Examples:
○ ALOHA:
■ Immediate packet transmission;
■ collisions detected via acknowledgments.
■ Acknowledgment absence triggers retransmission after random backoff.
■ Under light loads -> short access and txn delays
■ Under heavy loads -> no of collision ↑ -> ↓ throughput efficiency and ↑ txn delays
○ Slotted ALOHA:
■ Time is divided into slots(has max length packet);
■ nodes can only start transmissions at slot boundaries,
■ reducing collision probability due to synchronization and
■ enhancing throughput.
○ Carrier Sense Multiple Access (CSMA): Nodes listen to the medium(carrier
sensing)) and transmit when idle, or defer and retry based on backoff algorithms.
■ Non-persistent CSMA: Randomly waits for idle medium.
■ Persistent CSMA: Waits until ongoing transmission ends before accessing
the medium according to backoff algorithm.
■ p-persistent CSMA: Probability-based waiting for access to the medium.
○ IEEE 802.11 Distributed Coordination Function (DCF):
■ Operation:
■ A node picks a random value from the contention window and starts
a timer. The timer decrements after each slot; if another node starts
transmitting, the timer pauses and resumes after the current
transmission ends. The node transmits its frame when the timer
reaches zero.
■ Error Handling:
■ On transmission errors (e.g., missing acknowledgment), the
contention window size increases using a binary exponential backoff.
■
○ Efficiency:
■ CSMA protocols, like IEEE 802.11 DCF, have higher throughput efficiency
than ALOHA as they prevent the destruction of ongoing packets by detecting
potential collisions.
● Collision Management:
○ Hidden-Terminal Problem: Nodes cannot detect interference at the receiver, leading
to collisions.
○ Exposed-Terminal Problem: Nodes unnecessarily defer transmissions.
○ Solutions: Use of busy-tone solutions and RTS/CTS handshake to mitigate hidden-
terminal and exposed-terminal issues.
● Busy-Tone Solution:
○ Principle:
■ Uses two frequency channels—one for data and another for control signals.
Nodes emit an unmodulated wave on the control channel while receiving
packets to indicate busy status.
○ Transmission Protocol:
■ Nodes check the control channel for a busy tone before transmitting. If
detected, they defer using a nonpersistent CSMA-like algorithm. If clear, they
proceed with data transmission.
 ○ Advantages:
■ Solves both hidden-terminal and exposed-terminal problems by ensuring
nodes in the vicinity hear the busy tone.
■ Efficient use of bandwidth with minimal overhead on the control channel.
○ Challenges:
■ Requires precise frequency synchronization for effective operation.
■ Ensuring the busy tone's strength is optimal to prevent unnecessary
transmission suppression or signal destruction.
○ Variants:
■ PAMAS introduces a variant potentially offering enhancements or
adaptations to the basic busy-tone approach.
● RTS/CTS Handshake (IEEE 802.11):
○ Principle: Based on the MACAW protocol, it uses a single channel with two control
packets: Request To Send (RTS) and Clear To Send (CTS).
○ Operation:
■ Node B sends an RTS packet to Node C, indicating the transaction duration.
■ Upon receiving the RTS, Node C replies with a CTS packet, also indicating
the duration.
■ Node B transmits the data packet, followed by an acknowledgment from
Node C.
○ NAV(Network Allocation Vector) Mechanism: Nodes hearing RTS, CTS, data, or
acknowledgment set an internal timer (NAV) to avoid collisions during the
transmission duration. Nodes do not respond with CTS to avoid interference.
○ Collision Scenarios:
■ Nodes A and C can issue simultaneous RTS packets, leading to lost RTS but
no data frame collision.
■ Two issues in Figure 5.3:
● Left: Node D’s RTS to C collides with B’s CTS, causing C to fail in
decoding the duration field from A’s packet.
● Right: Node C’s RTS to D starts before sensing B’s CTS, leading to
improper decoding. Solution: Ensure CTS packets are longer than
RTS, allowing nodes to sense signals and defer transmissions.
○ Overhead Concern: The RTS/CTS handshake adds significant overhead (two
control packets plus acknowledgment). For small packets, the overhead may not
justify its use, whereas for long packets(hit by channel errors and need entire
retransmission), fragmentation (e.g., IEEE 802.11, S-MAC) and a single RTS/CTS
exchange for all fragments are recommended.
○
● Challenges: Overhead from control packets, potential collisions despite protocols, and
energy inefficiency with long packets.

)
(11.3.3 Further reading
General Literature on Mobile Ad Hoc Networking:

● Offers valuable inspiration despite some differences from sensor networks.

● Example: Cost-field-based approach (reference [910]).
○ Assigns a "height" to nodes representing the minimum energy required to reach the
source.
○ Similar to TORA ad hoc routing protocol [619], which assigns heights in hop counts.
○ Cost field updated by randomly delaying rebroadcasting of advertisement messages.

11.3.5 Further reading

● Routing and Topology Control:
○ Context: In clustered or dominating set-based networks, routing needs to consider
topology control and battery capacity.
○ References: [150, 356, 488, 628, 771]; LEACH [344, 346] also falls under this
category.
● Maximizing Data Flow for Multiple Source/Destination Pairs:
○ Objective: Optimize routes for multiple sources sending data to dedicated sinks,
maximizing network utility before energy depletion.
○ Reference: [781], which explores flow control algorithms applicable to similar
problems.
● Consider All Costs:
○ Insight: Path selection should consider all energy costs, including retransmission
costs due to residual error rates over wireless links.
○ Reference: Banerjee and Misra [51] compare local versus end-to-end retransmission
schemes.
● Integrate Scheduling and Power Control:
○ Approach: Cruz and Santhanam [185] and Bergamo et al. [65] integrate link
scheduling, power control, and routing to minimize average power consumption.
○ Complexity: These approaches involve complex optimization considering traffic
requirements for all links.
● Routing and Link Quality:
○ Challenge: Links vary in reliability and packet error rates, affecting routing protocol
efficiency.
○ Solution: Wang et al. [859] advocate for selecting neighbors based on reliable link
information provided by the link layer.
● Routing and Lifetime Guarantees:
○ Objective: Ensure network lifetime through routing decisions.
○ References: [716, 717] attempt to provide guarantees on the network's operational
lifetime.
● Routing for One-Shot Queries:
○ Scenario: Handling queries and routing answers back without relying on outdated
topology information.
○ Approach: Helmy [348] proposes a scheme using R-hop neighborhoods and contact
nodes for queries outside immediate vicinity.

These topics provide deeper insights into various aspects of unicast routing in ad hoc and wireless
sensor networks, highlighting different optimization objectives, challenges, and solutions. )

Data Management in Sensor Networks

Objective:
The goal of data management in sensor networks is to abstract the logical view of data (naming,
accessing, operations) from its physical form. Users should focus on queries without worrying about
the sensor network's intricacies. Essentially, a sensor network's data management system functions
like a distributed database but differs significantly from traditional ones.

Key Differences Between Sensor Network Data Management and Traditional

Distributed Database Systems:

1. Network Environment:
○ Traditional Systems: Data management is independent of network specifics.
○ Sensor Networks: Must consider network details since sensors can fail, have limited
storage, computing power, and battery life, and offer weak network services.
2. Data Characteristics:
○ Traditional Systems: Manage finite, static datasets.
○ Sensor Networks: Handle infinite data streams from sensors, requiring new
techniques for processing and analyzing these continuous data streams using limited
resources.
3. Data Accuracy:
○ Traditional Systems: Data is generally accurate.
○ Sensor Networks: Sensor data can be erroneous, described by probability
distributions. Systems must handle and correct these errors to provide reliable data.
4. Power Constraints:
○ Traditional Systems: Less emphasis on power usage.
○ Sensor Networks: Must minimize power consumption to extend network life,
prioritizing local data processing to reduce transmission costs.
5. Query Types:
○ Traditional Systems: Handle finite, predefined queries.
○ Sensor Networks: Support long-running queries (monitoring over time) and ad-hoc
queries (current status). Must balance query accuracy with resource usage,
employing techniques like random sampling to reduce costs.
6. Query Processing Techniques:
○ Traditional Systems: Rely on fixed cost models and statistical information.
○ Sensor Networks: Lack reliable statistical data, making it hard to predict sensor data
behavior. Queries must minimize power consumption and adapt to data stream
dynamics. Traditional locking mechanisms and fixed databases don't suit the
uncertain, infinite data streams of sensor networks, requiring real-time processing
and approximate query responses.
7. Data Volume and Storage:
○ Traditional Systems: Can store large datasets centrally.
○ Sensor Networks: Generate large data volumes that can't all be stored. Centralized
storage isn't feasible due to bandwidth, processing, and power limitations. Instead,
they use distributed, in-network processing to enhance performance and fault
tolerance.

Conclusion
Data management in sensor networks involves handling continuous, potentially erroneous data
streams within strict power and resource constraints, unlike traditional database systems. This
requires innovative approaches to query processing, error handling, and data storage, aiming for
efficiency and real-time response while maintaining accuracy and extending network life.

Architecture of Data Management Systems in Sensor Networks

Data management systems in sensor networks can be categorized into four models: centralized,
semi-distributed, distributed, and hierarchical. Each model has unique characteristics and approaches
to managing data from sensor networks.
2.1 Centralized Model

In the centralized model, the process involves two steps:

1. Data Extraction: Sensor data is collected and sent to a central server.

2. Query Processing: The central server stores the data in a database and processes queries.

Advantages:

● Simplicity in design and implementation.

Disadvantages:

● The central server is a performance bottleneck and a single point of failure.

● High communication costs as all sensors transmit data to the central server.

2.2 Semi-Distributed Model

In the semi-distributed model, sensors perform some data processing locally before transmitting data.
This reduces communication overhead and distributes the computational load.

Examples:

1. Fjord:
○ Adaptive Dataflow System: Combines push (sensor data) and pull (non-sensor
data) mechanisms for query processing.
○ Components:
■ Adaptive Query Processing Engine: Adjusts execution plans based on the
computing environment.
■ Sensor Proxy: Acts as an interface between sensors and the query
processor, managing data delivery and local computations.
2. Cougar:
○ In-Network Aggregation: Sensors perform local computations and communicate
with nearby sensors to aggregate data, reducing the need to send all data to a central
server.
○ Flexible Query Processing: Only relevant data is extracted based on query
workload, optimizing resource use.

2.3 Distributed Model

In the distributed model, each sensor has substantial storage, computation, and communication
capabilities. Data processing and storage are fully distributed among sensor nodes.

Key Technique:

● Distributed Hash Table (DHT):

○ Events are hashed and stored at the sensor node closest to the hash value.
○ Queries are processed by sending them to the corresponding node based on the
hash value.

Disadvantages:

● Assumes sensors have high capabilities, similar to standard computers.

● Suitable mainly for key-based queries, leading to high communication costs.

2.4 Hierarchical Model

The hierarchical model introduces a two-layer system: sensor network layer and proxy network layer.
Layers:

1. Sensor Network Layer:

○ Sensors perform local computations, receive commands from proxies, and deliver
processed data to proxies.
2. Proxy Network Layer:
○ Proxies have higher computational and storage capabilities.
○ Functions include receiving queries, issuing control commands, processing data, and
delivering results.

Advantages:

● Distributes computation and communication loads among proxies, improving efficiency and
fault tolerance.

Data Model and Query Language in Sensor Networks

3.1 Data Model

Sensor network data models extend traditional database models to accommodate the unique
characteristics of sensor data.

Examples:

1. TinyDB:
○ Extended Relational Model: Treats sensed data as a single, infinite logical table
with attributes for perceptive data (e.g., temperature) and sensor characteristics (e.g.,
sensorID).
○ Operations: Extend traditional relational algebra to handle infinite relational tables.
2. Cougar:
○ Hybrid Model: Combines relational data (stored data) and time-series data (real-time
sensor data).
○ Operations: Include relational algebraic operations and time-series operations (e.g.,
projection, product, aggregation).

3.2 Query Language

Query languages in sensor networks need to handle real-time, periodic, and uncertain data efficiently.

Examples:

1. TinyDB (TinySQL):
○ Based on SQL, it supports:
■ Selection and projection.
■ Determining sampling rates.
■ Group aggregation and user-defined aggregation.
■ Event triggers, lifetime queries, setting storage points, and simple joins.

Key Characteristics:

● Queries can be snapshot, continuous, event-based, lifecycle-based, or accuracy-based,

addressing the diverse needs of sensor network applications.

Summary
The data management architecture in sensor networks ranges from simple centralized models to
complex hierarchical models, each balancing trade-offs between communication cost, computational
load, and fault tolerance. The data models and query languages in these systems are adapted to
handle the continuous, real-time, and uncertain nature of sensor data, ensuring efficient data
processing and query execution.

Example Unicast Protocols for Energy-Efficient Routing

Attracting Routes by Redirecting

Gomez et al. proposed a method where nodes overhear packets and, if a node can offer a more
energy-efficient route, it sends route redirect messages to improve the path. This method reduces
administrative overhead but relies heavily on nodes overhearing traffic, which is less suitable for
WSNs.

Distance Vector Routing on Top of Topology Control

A Bellman–Ford-type algorithm is used to find paths with minimal power consumption, leveraging the
relay regions concept to create energy-efficient routing paths.

Maximizing Time to First Node Outage as a Flow Problem

Chang and Tassiulas use a centralized approach to maximize the time until the first node depletes its
battery. They model the problem as a linear programming issue, taking into account both energy costs
and battery capacities, and propose two distributed approximation algorithms. This approach can
extend system lifetime by up to 60% compared to minimum energy routing.

Maximizing Time to First Node Outage by a Max–Min Optimization

Li et al. present a max–min optimization approach, proposing two algorithms:

1. Max Min zPmin: Chooses paths based on remaining battery power, limiting maximum power
consumption to a factor z of the most efficient path.
2. Zone Routing: Divides the network into zones, with routing handled locally within zones,
reducing the need for global battery level knowledge.

Maximizing Number of Messages

Kar et al. aim to maximize the number of messages sent before the network runs out of energy. They
introduce the CMAX algorithm, which selects paths based on link weights determined by energy costs
and battery usage. This algorithm performs well even without admission control and can outperform
specialized algorithms in terms of network lifetime.

Bounding the Difference Between Routing Protocols

Alonso et al. analyze the potential impact of routing protocol choices on energy efficiency. They show
that in networks where most nodes are more than one hop away from the base station, the worst
possible routing protocol will only reduce energy efficiency by a factor of up to 2|S1| - 1 compared to
the best possible routing. This indicates that the choice of routing protocol has limited but significant
practical impact on energy efficiency.

11.5.3 Further reading on geographic routing

1. Impact of Localization Errors:
○ He et al. [339]: Investigate the impact of localization errors on routing protocols,
finding that certain protocols maintain acceptable delivery ratios and path length
overhead within 40% localization error.
 ○ Seada et al. [748]: Discuss the susceptibility of face routing protocols to localization
errors and propose improvements in constructing planar graphs.
○ Kim et al. [419]: Provide additional results on the impact of localization errors on
routing protocols.
2. Location Services:
○ Basagni et al. [61]: Propose the DREAM protocol, which optimizes location updates
based on node mobility and distance effects.
○ Basagni et al. [60]: Highlight the high accuracy of distributed position tables
achieved through the DREAM protocol.
○ Li et al. [483]: Describe a scalable, distributed position database using consistent
hashing and minimal updates.
○ Grossglauser and Vetterli [306]: Discuss a scheme where node mobility updates
location information without additional communication overhead.
3. Location-Aided Routing (LAR):
○ [423]: Similar to LBM, LAR uses location information to assist in flooding phases of
ad hoc routing protocols, utilizing expected and request zones for route requests.
4. Making Geocasting Energy Aware:
○ GEAR [919]: Introduces load-splitting among neighbors during geocasting to equalize
energy consumption, leading to extended network lifetime and better post-partition
connectivity.
5. Geographic Routing Without Geographic Coordinates:
○ Rao et al. [681]: Introduce virtual coordinates for geographic routing, where nodes
estimate their coordinates iteratively based on neighbor averages without relying on
actual physical locations.
○ Virtual coordinates are shown to be effective even when perimeter nodes do not
know their physical location.
6. Link Asymmetry:
○ Zhou et al. [934]: Demonstrate that geographic routing performance degrades
significantly in the presence of link asymmetry, highlighting challenges in routing
under such conditions.

These topics illustrate ongoing research efforts to improve geographic routing protocols in wireless
networks, addressing issues such as localization errors, energy efficiency, virtual coordinates, and link
asymmetry. Each study contributes valuable insights and proposed solutions to enhance the
robustness and efficiency of geographic routing protocols in various network environments.