Using Robot Mobility to Exploit Multipath Fading Magnus Lindhé
Karl Henrik Johansson
ACCESS Linnaeus Centre
Royal Institute of Technology
Stockholm, Sweden
Using Robot Mobility to Exploit Multipath Fading Magnus Lindhé
Karl Henrik Johansson
ACCESS Linnaeus Centre
Royal Institute of Technology
Stockholm, Sweden
Motivating scenario
Follow pre-defined path while ensuring communication QoS Received signal strength (dBm)
Problem formulation
Multipath Fading
No line-of-sight: Rayleigh fading
SNR is exponentially distributed
Key idea
Pr (dBm)
Three strategies
Strategy 0: No stopping
Follow the reference trajectory, no feedback from radio.
Strategy 1: Periodic stopping
Drive a constant time, then stop and measure SNR.
Stop duration depends on SNR.
Strategy 2: Controlled stopping
Stop ”on demand”, when buffer is filling up and we’re not too far behind.
Strategy 0: No Stopping
Strategy 0: No Stopping
Strategy 1: Periodic stopping
Strategy 1: Periodic stopping
Strategy 2: Controlled Stopping
Strategy 2: Controlled Stopping
Comparison
No stopping:
+ Perfect reference tracking
+ Low energy consumption
- No improvement in link capacity
Periodic stopping:
+ Constant capacity improvement
+ Simple control law
- Stop&go is energy-consuming
- No feedback from lag or buffer size
Controlled stopping:
+ Adaptively trades capacity for reference tracking
+ Energy-conscious
- More complex control law
Expected link capacity
Experiment
Preliminary results show improvement in link capacity
Future: integrate in KTH testbed
KTH Wireless Sensor Network Testbed
Wired power supply, wireless programming and communication
Ultrasound localization
Robot
Summary
Periodic stopping Constant drive time,
SNR-dependent stop time Controlled stopping Minimize tracking error, buffer
size and power consumption. Multipath fading causes fast variations in SNR and link capacity – sometimes a gain!
Adapt robot motion to exploit positions with good signal strength. Increases average capacity.
Two strategies:
Improved link capacity:
Preliminary experiments
Future research
Design outer feedback loop that estimates mean SNR:
Comments