Open Source UAV and mobile cellular networks

| | Comments ()

The safe operation of an UAV requires a communication link to handle telemetry data, control commands and other information between the vehicle and the ground control station (GCS).

A simple and affordable way to overcome the range limitation is running the UAVs missions over the mobile radio infrastructure.

This blog post contains technical notes, references and pointers on Open Source UAVs and the feasiblity of these communication links from a vehicle's command and control perspective.

This is the third entry on Open Source UAVs. The previous entries are available here and here.

Cellular networks

A cellular network is a communication network where the last link is wireless. The network is distributed over land areas called cells, each served by at least one fixed-location transceiver.

The base stations provide the cell with the network coverage which can be used for transmission of voice, data and others.

A cell typically uses a different set of frequencies from neighboring cells, to avoid interference and provide guaranteed QoS within each cell. This division into cells follows a pattern which depends on terrain and the reception characteristics.

The cells are assigned with multiple frequencies which have corresponding radio base stations. The frequencies can be reused in other cells, provided that the same frequencies are not reused in adjacent neighboring cells as that would cause co-channel interference.

As the distributed mobile units move from cell to cell during an ongoing continuous communication, the cell-to-cell frequency switching is done electronically without interruption and without a base station operator or manual switching. This is called the handover or handoff.

The new channel is automatically selected for the mobile unit on the new base station which will serve it.

The mobile phone network

The most common example of a cellular network is a mobile phone network or cell phone network.

A mobile phone is a portable telephone which receives or makes calls through a cell site (base station), or transmitting tower. The radio waves are used to transfer signals to and from the cell phone.

Large geographic areas are split into smaller cells to avoid line-of-sight signal loss and to support a large number of active phones in that area. All of these cells are connected to telephone exchanges or switches, which in turn connect to the public telephone network.

A cellular network is used by the mobile phone operator to achieve both coverage and capacity for their subscribers. The cells usually have a range of 0.80 km (city) to 8 km (rural areas).

The following picture comes from OpenSignal. It shows the global state of the mobile network with 3G or better availability in August 2016.

This mobile network infrastructure allows mobile units to be connected to the public switched telephone network and public Internet.

The 4G LTE network architecture model

The usual mobile-radio network architecture consists of the following blocks:

  • A network of radio base stations forming the base station subsystem.
  • The core circuit switched network for handling voice calls and text.
  • A packet switched network for handling mobile data.

We will use the 4G LTE network architecture as the reference model in this post. If you are not familiar with it, you may find interesting this quick overview on the the three main components: the User Equipment (UE), the Evolved UMTS Terrestrial Radio Access Network (E-UTRAN) and the Evolved Packet Core (EPC).

I would also suggest to have a look in this article. It describes the LTE protocol stacks (user and control planes) and it illustrates the traffic flow on the LTE network in great detail.

Cellular for the skies

The 'Cellular for the Skies: Exploiting Mobile Network Infrastructure for Low Altitude Air-to-Ground Communications' article is an useful and condensed study of the problem from the UAS communication perspective and civil applications.

As described in the article, and as supported by the theoretical data rates and latencies for mobile cellular technologies showed in the table, the current infrastructure has the capabilities in terms of throughtput and latencies to support UAV operations close to 'real-time'.

Technology Downlink (Mbps) Uplink (Mbps) Latency (ms)
EDGE 0.236 0.059 150
UMTS 2 0.384 100
HSPA+ 42 11.5 50
LTE 300 75 10
LTE-A 3000 1500 10

These capabilities can make mobile cellular technology suitable not only for telemetry and commands but also for streaming video in private and/or public networks, including Internet.

Taking into account the cellular networks were not designed to support air-to-ground communications the UAV missions are sensitive to issues such as:

  • possible loss of radio coverage even at low altitudes due to base station antennas often tilted down
  • disaster situations where the infrastructure can fail to operate
  • geographic regions where the coverage is not available
  • packet-switched networks delays
  • ...

To overcome these issues related to performance/latency, packet loss, etc. the UAVs can always fallback on fail-safe mechanisms switching from semi-autonomous operations to autonomous flight modes.

Some fail-safe strategies in front of a permanent loss of connectivity and link could be flying to the last geolocation with radio coverage, emergency landing, path planning to keep continuous communication, etc.

Related to UAV's monitoring and controlling protocols we are interested on MAVLink. This protocol defines the telemetry messages as arrays of 8 to 263 bytes, transmitted with a frequency of up to 20 Hz. So the telemetry streams will require bitrates up to 42 kbps.

In the case of video streaming, the bitrate required to receive video with acceptable quality lay between 200 kbps and 800 kbps, depending on the codec used, frames per second and video bitrate.

Finally the article also contains an experimental setup and results section where the authors assess the use of mobile networks with UAVs in order to conclude the measured latencies and jitter could be considered excessive for manual control, especially in the case of HSPA+ and EDGE, but they are perfectly suitable for semi-autonomous flight modes.

Wireless 'modem' USB sticks

Setting up Internet access over a 4G LTE network on the Open Source flight platform is straight forward.

On our case the installation only requires plugging a 4G LTE USB dongle in the companion computer and then configuring the new wireless interface via the USB subsystem.

These dongles plays the hardware role of the User Equipment in the 4G LTE network. The dongle contains the modules handling the communication functions, the data streams termination and the Universal Integrated Circuit Card (UICC) module, also known as the SIM card for LTE equipments.

The UICC runs an application known as the Universal Subscriber Identity Module or USIM. The USIM stores user-specific data such as the user's phone number, the home network identity, the security keys, etc.

There is good technical documentation covering the installation and configuration of specific hardware. Take as example this technical documentation to set up a Huawei E3276-150 4G/LTE USB modem quickly. Other modems/models/versions should need a similar installation/configuration.

The navigation grid

From a command and control perspective we can consider our vehicle a flying computer over a non reliable communication network where the communication link is independent of the kind of mission the vehicle is running.

In 'simple' scenarios the vehicle can be pre-programmed with well-known coordinates. In those scenarios the vehicle can switch between autonomous and semi-autonomous flight modes depending on the state of the communication link/network.

On more complex scenarios we will be interested to run some kind of 'smart missions' where the next waypoint is evaluated on the current one (or previous ones) depending on the fixed and dynamic mission factors.

The fixed factors here could be the cellular base station antennas location, the terrain characteristics, the legal aspects and regulation, etc.

While the dynamic factors could be the weather, the temperature, some non expected software/hardware issue, an updated goal, etc.

Beyond of the concrete model used and the responsibility of the vehicle to take these decisions locally on its own or over a remote and/or more centralized architecture, a 'navigation grid' as a data structure could be used to support these 'smart missions'.

This navigation grid would catch the required fixed and dynamic mission factors in order to support the navigation of the vehicle. Those factors can be included in the grid as normalized and adapted weights to be used with the different pathfinding and graph traversal algorithms.

One simple example of grid used by these 'navigation algorithms' could be the next picture. It displays an efficiently directed path between the green node and the blue node plotted over a possible 'navigation grid' using the classic A* algorithm.

In this case, the navigation grid contains an unique weight per square although it is just a matter of design. Several weights could be calculated with different algorithms and priorization criterias if needed.

Take into account it depends on the navigation grid design what a weight represents and how it can be influenced via its components and factors.

It is also important mentioning the navigation grid can be used along the mission to track/discover non available information when the mission started. This new information, real or estimate, can be used to run more efficient navigation algorithms (distance, power, etc) available in the navigation stack but disabled while this information was not available.

Wrapping up

This blog post explores the feasibility of supporting autonomous and semi-autonomous flight modes over mobile cellular networks from a technical point of view. It comments on the basic concepts, major blocks and required interfaces to understand the advantages and disadvantages of these networks.

The numbers behind the different available sources and the documented tests support the fact of relying on this kind of networks to command and control civil UAVs if needed.

While the advantages of this kind of infrastructure (coverage, cost, operational range, etc) are clear in the UAV domain, the entry also identifies and comments on non obvious 'constraints by design', like the maximum altitude limitation due to base station antennas often tilted down.

Related to MAVLink, the communication protocol used by our flight platform, and the chance to stream video; the entry covers the conditions, configurations and thresholds where the bandwidth and latency require special consideration.

The blog post also explores the required hardware to connect the vehicle over a 4G LTE network. This kind of hardware, based on an affordable USB dongle, is well supported in GNU/Linux, the operating system used in our companion computer.

Finally the 'navigation grid' is included as part of the 'smart' mission concept, and how this data structure can be used to enable a more flexible and dynamic navigation along a mission.

It is expected the future enhancements for 4G networks and evolution to 5G will benefit UAV communications even further.


comments powered by Disqus

Recent Entries