Wireless feedback control loops with neural networks to predict target system states

At least one non-transitory computer readable medium comprising computer readable instructions which, when executed, cause at least one processor to at least:

• execute a first neural network and a second neural network to predict a future value of a state of a target system corresponding to a future time relative to a prior time associated with a first measurement of the target system obtained via a first wireless link, a time interval between the prior time and the future time based on a first latency associated with the first wireless link and a second latency associated with a second wireless link, the first neural network to predict an initial value of the state of the target system based on the first measurement and a first one of a sequence of prior values of a control signal previously generated to control the target system during the time interval between the prior time and the future time, the second neural network to predict intermediate values and the future value of the state of the target system iteratively based on the initial value predicted by the first neural network and remaining ones of the sequence of prior values of the control signal, the second neural network to operate at a faster rate than the first neural network to cause the second neural network to perform a first number of predictor prediction iterations during a time interval in which the first neural network is to perform a single observer prediction iteration, the first number of predictor prediction iterations based on a sum of the first latency and the second latency;
• and
• output the predicted future value of the state of the target system to a controller that is to generate a new value of the control signal that is to be sent to the target system via the second wireless link.

A wireless feedback control method comprising:

• accessing a first measurement of a target system received via a first wireless link;
• predicting, with a first neural network and a second neural network, a future value of a state of the target system corresponding to a future time relative to a prior time associated with the first measurement, a time interval between the prior time and the future time based on a first latency associated with the first wireless link and a second latency associated with a second wireless link, the predicting of the future value of the state of the target system including (i) predicting, with the first neural network, an initial value of the state of the target system based on the first measurement and a first one of a sequence of prior values of a control signal previously generated to control the target system during the time interval between the prior time and the future time, and (ii) predicting, with the second neural network, intermediate values and the future value of the state of the target system iteratively based on the initial value predicted by the first neural network and remaining ones of the sequence of prior values of the control signal, the second neural network to operate at a faster rate than the first neural network to cause the second neural network to perform a first number of predictor prediction iterations during a time interval in which the first neural network is to perform a single observer prediction iteration, the first number of predictor prediction iterations based on a sum of the first latency and the second latency;
• outputting the predicted future value of the state of the target system to a controller that is to generate a new value of the control signal;
• and
• sending the new value of the control signal to the target system via the second wireless link.

A wireless feedback control system comprising:

• means for receiving a first measurement of a target system via a first wireless link;
• means for predicting a future value of a state of the target system corresponding to a future time relative to a prior time associated with the first measurement, a time interval between the prior time and the future time based on a first latency associated with the first wireless link and a second latency associated with a second wireless link, the means for predicting including (i) first neural network means for predicting an initial value of the state of the target system based on the first measurement and a first one of a sequence of prior values of a control signal previously generated to control the target system during the time interval between the prior time and the future time, and (ii) second neural network means for predicting intermediate values and the future value of the state of the target system iteratively based on the initial value predicted by the first neural network means and remaining ones of the sequence of prior values of the control signal, the second neural network means to operate at a faster rate than the first neural network means to cause the second neural network means to perform a first number of predictor prediction iterations during a time interval in which the first neural network means is to perform a single observer prediction iteration, the first number of predictor prediction iterations based on a sum of the first latency and the second latency, the means for predicting to output the predicted future value of the state of the target system to a controller to generate a new value of the control signal;
• and
• means for transmitting the new value of the control signal to the target system via the second wireless link.