CarderPlanet
Professional
Scientists have presented a "cognitive radio" with a neural network that analyzes changes in the environment and adjusts its settings accordingly. This makes the new radio more robust in unknown environments, such as space, far beyond Earth's orbit.
Conventional radio communications in space are influenced by many factors: Earth's oscillating ionosphere, low-level radio emissions, microwave background radiation. All of them can suppress the signal. However, for future space missions, a stable connection needs to be ensured. For this, the communication system must learn to adapt to the environment.
For example, astronauts on Mars would use a repeater satellite orbiting the planet to connect to a ground station on Earth. As the space environment changes, the radio settings on the ground station, the satellite orbiting Mars, and the Martian spacecraft will need constant adjustments. As a result, astronauts will have to wait 8 to 40 minutes to receive instructions or other information.
Worcester Polytechnic Institute and Penn State, in collaboration with NASA, recently tested the first cognitive radios designed to operate in space and keep missions in contact with Earth. They provided a clear signal for communication with the International Space Station (ISS).
The idea of such a radio was first voiced by Joseph Mitola III at the Royal Institute of Technology in Stockholm in 1998.
The difference between cognitive radio is that its neural network learns from data from the environment itself, and not from a mathematical model. It receives information about which signal modulation works best or which frequencies travel farthest, and processes this data to determine which radio settings are needed for optimal communication. A key feature of a neural network is that it can optimize the relationship between input and output over time. In noisy environments where the signal does not pass, the radio may first try to increase the transmit power. It will then determine if the received signal is clearer; if so, the radio will increase its transmit power to see if it improves reception. But, if the signal does not improve, the radio station may try a different approach, such as changing frequencies.
Cognitive radio uses a wireless system called software-defined radio to control basic settings. The main functions that are implemented using hardware in a conventional radio are performed here using software, including filtering, amplification and signal detection.
So far, experiments with cognitive radio are still limited in scope. At their core, neural networks are complex algorithms that require huge amounts of data to operate. They also require a lot of computing power. The radio equipment must be designed with sufficient flexibility to adapt to the conclusions received. And any successful cognitive radio needs to make these components work together.
NASA's Glenn Research Center has created the SCaN test bed specifically to study the use of software-defined radios in space. It was launched by the Japan Aerospace Exploration Agency and installed on the ISS main lattice frame in July 2012. Prior to its decommissioning in June 2019, SCaN allowed researchers to test how well software-defined radios could meet anticipated requirements in space, such as real-time reconfiguration for orbital operations, development and validation of new software for non-standard space networking and cognitive communications. ...
The test bench consisted of three software-defined radios broadcasting in the S-band (2 to 4 GHz) and Ka-band (26.5 to 40 GHz) and receiving in the L-band (1 to 2 GHz). The SCaN test bed could communicate with NASA's satellite tracking and relaying system in low Earth orbit and a ground station at the Glenn Research Center in Cleveland.
The cognitive radio at the ground station made a decision about the "action" or set of operating parameters for the radio, which it sent to the test bed transmitter and two modems. The action contained a specific baud rate, modulation scheme, and power level for the test bench transmitter and ground station modems. The first tests were completed in May 2017. The radio system operated under dynamic and challenging communication conditions, including fluctuations in the atmosphere and weather. Oftentimes, solar panels and other protrusions on the ISS created a large number of echoes and reflections that this system had to account for.
During each pass of the ISS, the neural network compares the quality of the communication line with the data of previous passes. It then selects the previous pass with the conditions that are most similar to the current one as the starting point for the setup. After that, the neural network adjusted the radio parameters in accordance with the conditions of the current passage. These settings included all the elements of the wireless signal, including baud rate and modulation. If using just one pass was not enough, the system created an individual solution from several previous passes.
The radio autonomously selected the settings to avoid losing contact, and the connection remained stable. It also had sufficient signal strength to send data.
However, this experiment also revealed several problems that need to be addressed before using the new radios.
The biggest problem has been called "catastrophic oblivion." This happens when the neural network receives too much new information too quickly and therefore forgets a lot of what it has already learned. In such situations, the capabilities of cognitive radio were significantly impaired. Then the scientists decided to implement ensemble learning - an experimental technique that uses a set of neural networks, each responsible for learning under a limited set of conditions - in this case, on a specific type of communication channel. The meta-neural network decides which networks to use in the current situation.
After introducing this type of training in August 2018, researchers found that the incidence of catastrophic oblivion decreased. However, questions remain for the meta-neural network itself, for example, how to train it to choose the best available networks for a specific scenario.
To further demonstrate the capabilities of cognitive space communications, NASA plans to launch a constellation of three cubesats in the next few years. They intend to use it as a relay system to figure out how several cognitive radio stations can work together.
