Wireless Communication Test Bed
Experiment 0
Report on the March 7, 2004 Experiment
Version 0.2
2 April 2004
Prepared Under Subcontract SC03-034-191 with L-3 ComCept
Prepared by:
Timothy X. Brown, University of Colorado at Boulder
303-492-1630
Brian M. Argrow, University of Colorado at Boulder
303-492-5312
1.INTRODUCTION
This document describes the Wireless Communication Test Bed Experiment 0 which took place on March 7, 2004. Experiment 0 was the first combined test of the different test bed components and the main goal was to ascertain any operational issues. The following tests were performed:
- UAV mesh radio node (MRN) and flight control radio interference test;
- 802.11b radio range between ground nodes;
- Mesh network automatic formation;
- Monitoring data collection; and
- Mesh network throughput.
In summary, the Experiment successfully gathered data useful for future experiments and deployments. The UAV control and MRN did not interfere, radio ranges of 1.2km were demonstrated between ground nodes; the network always formed; the monitoring collected data;but, we did not measure throughput. Issues uncovered include a routing software error, and a need for better experimentation documentation.
The rest of the document summarizes the test procedures and results, issues uncovered and issues that remain to be resolved.
2.Tests
The experiments took place at the Table Mountain (TM) test site. The personnel met at 12pm at the B9 building and network setup began immediately. A list of personnel is in Appendix A. A timeline of activities is in Appendix B. The laptop gateway was connected to the TM network and connectivity to the CU monitoring server was established. Two MRN were placed outdoors on top of the B9 building (Note that the B9 building is located on the flank of the mountain so that its flat roof is level with and abuts with the top level of TM). One was placed on the ground, and the other was placed on a ladder 2m above the ground level. Power was provided via the power over Ethernet cables from the building. Only power and not Ethernet was applied to the cables. A third radio was mounted on a vehicle roof approximately 1.5m above the ground level. Power was provided by a cigarette adapter. A fourth radio was mounted on the UAV and powered via aircraft onboard battery.
2.1.Test 1: UAV Interference
For the initial flight tests the system used to control the AUGNet aircraft is a 72 MHz R/C transmitter and receiver system from Futaba. Because this system can also be used as a backup method for operations within the AUGNet experiments, it is of interest to verify any bad interactions between the R/C receiver and the 2.4 GHz 802.11 payload system.Standard qualitative tests, as specified by Futaba [1], were performed to determine if there where any loses in radio range for the R/C system.
A standard test performed on R/C radio systems to verify their range is to place the model on the ground, with the receiver antenna properly mounted, and to walk away from the aircraft with the transmitter antenna collapsed, varying the control input. A spotter at the aircraft watches and listens to determine when the servos in the aircraft begin to jitter and finally stop responding. If the point at which the servos stop responding is greater than 100 feet, than the radio system is determined to have significant range for standard R/C flight operations [1].
This test was performed with varying locations of the receiver antenna, while the 802.11 radio was operating in an active ad hoc network. It was found from these tests that the 802.11 radio system gave no noticeable decrease upon the radio range and all tests provided a ground range test of more than 150 feet. However by varying the antenna location, it was found that the payload computer, which operates around 100 MHz, decreased the range when the receiver antenna was placed within close proximity to the CPU. Though the reduced range was not less than 100 feet, to ensure sufficient radio range for the R/C system the receiver antenna should be placed sufficiently far from the payload computer. From these tests, more than four inches of separation of the receiver and computer is sufficient.
Figure 1: Interference Testing
Future tests will look to determine if a running engine will have any affect on the radio systems in the airplane. Futaba specifies that up to a 10% loss due to the engine is acceptable under normal operations. Thus, a ground range test up to 15 feet can be accepted.
2.2.Test 2: 802.11 Range Between Ground Nodes
A primary design factor for the network is the range between the ground nodes. This test is complicated by the fact that reception of packets is stochastic with some packets being lost even if they radios are close and some packets being received even if the radios are distant. We attempted two versions of range test to see what would work best. In both cases, two nodes were placed at building B9, one at ground level and a second at 2m above ground level as previously described.
Figure 2: Range Test
In the first test, a node was mounted on a vehicle approximately 1.5m above the ground level. This vehicle was driven different distances from the test location. The vehicle node was pinged from the gateway node at B9 and connectivity was measured. The path of the ping packets was from the gateway to one of the selected nodes outside B9 to the vehicle node. This proved unsatisfactory due to a problem that was found in the software. The GPS was found to delay packets for up to one second. In the two radio hops this delay led to irregular behavior that prevented consistent measurements and even stopped one of the nodes at one point.[1]
The second test was simpler. Each node periodically sends out a beacon packet as part of the 802.11 MAC ad hoc operation. The Berkeley Variatronics Yellow Jacket 802.11 LAN Analyzer was used as the mobile receiver. This was held at 2m above ground level and the reception of the beacon packets was observed. Further, the received signal power of each received packet from the transmitter was graphed and the average received signal level at each location was recorded.
The results are shown in Table 1and plotted inFigure 3. From this data we see that the range is about 400m when the node is placed on the ground and 1200m when placed 2m above the ground. Figure 4: Indicates some of the places where range measurements were made to give an idea of what these ranges mean on the test bed.
The two-ray ground model [2] indicates that the received signal power is given by where K is some constant, h is the height of the transmitter, and d is the transmitter receiver separation. If we require some specific received power, we can solve for the distance where c is some constant. Thus a factor of 10 increase in transmitter height will yield an approximately factor of 3 increase in range. This is observed in the range experiment where a factor of 10 height increase from 0.2 to 2m yields a factor of 3 range increase from 400 to 1200m. Extrapolating this data (for purposes of estimating future performance) to the UAV operating height of 500ft (150m) and assuming that all other factors are equal suggests an approximately 10km range from UAV to ground.
Table 1: The received signal strength (RSSI) at different heights
and distances to a transmitter located 2m above the ground
Distance (m) / RSS (dBm) / Distance (m) / RSS (dBm)
34 / -40 / 48 / -40
45 / -50 / 123 / -50
64 / -60 / 264 / -60
162 / -70 / 487 / -70
317 / -80 / 673 / -71
410 / -83 / 760 / -75
954 / -79
1100 / -80
1200 / -83
1500m / Signal lost
Figure 3: Plot of range data
Figure 4: Experimental Layout
2.3.Test 3: Mesh Network Connectivity
The mesh network connectivity was tested simply by powering up the MNRs in different configurations and measuring if nodes were able to reach each other when pinged from the gateway. Connectivity could be either direct or via multihop routing. In all cases tested connectivity was established. Some of these cases included routing through the MNR mounted in the UAV.
2.4.Test 4: Monitoring Data Collection
This test showed whether the monitoring software on the MNR was able to collect data and send it to the monitoring server, aka Morse. Morse was receiving packets during the majority of all experiments. Collection was only stopped during idle periods between experiments. Data was correctly parsed and stored in the database. Statistics such as packets received, dropped, and sent all appear to be reasonable and accurate. The route information stored in the database makes it possible to see a correlation between the times when multi-hopping routes are active/inactive and when nodes are connected/disconnected (for example, it is possible to see the time when the fixed node stopped communicating).
The only thing not quite accurate in the database has to do with GPS. It is hard to see whether or not latitude/longitude values are accurate because no frame of reference was collected at the test site. This will be collected at a future test.
2.5.Test 5: Throughput
In the throughput test, packets were sent at increasing rates via ping and the packet success rate observed. Unfortunately, due to the interaction with the GPS, no meaningful measurements were made. It is expected that future experiments will be able to make this measurement without any problem.
3.Issues
The experiments revealed an error with our GPS coordinate recording software. The software would cause the MRN to hang in a wait state for up to one second stopping all other activity on the MRN. The error was due to some last minute code modifications that were not tested in the lab. The error has since been fixed. Future experiments will require that all software must be tested in the lab before being fielded.
Another general problem was the documentation of experiment activities. This became especially acute as some experiments did not work as planned and testing deviated from the plan. Experimental procedures and checkout sheets have been developed with areas for documenting planned activities, issues, planned and actual start/stop times, and experimental check out lists.
There are some remaining issues. Tests with a flying UAV need to be performed. The monitoring interface needs finalized and remote access made available. The scripts for other tests need to be written. These issues are being addressed in a timely manner.
4.References
[1] Futaba Frequently Asked Questions, Range Testing Your Futaba R/C Aircraft System,
[2] Rappaport, T., Wireless Communications Principles and Practice, 2nd Ed. Prentice Hall, 2000.
Appendix A:List of Personnel
Experiment personnel included:
Brian Argrow
Tom Bateman
Tim Brown
Corey Dixon
Sheetalkumar Doshi
Sushant Jadhav
Jay Jones
Roshan Thekkekunal
Appendix B:Experiment Timeline
The timeline for the experiment was:
12:00 Personnel begin arrival at building B9 at test site
13:00 Network connectivity established, outdoor fixed nodes installed.
13:30 Node attached to vehicle, begin throughput test
14:30 UAV interference test.
15:00 Range measurements
16:00 Begin packing
16:30 Site departure
1
[1] This problem with the GPS has been resolved since Experiment 0.