Cognitive Pilot Develops Voice AI Assistant for Driverless Mini Tractor
The system continues to operate even when the connection is unstable.
Cognitive Pilot has developed a voice AI assistant for an autonomous mini tractor. The system is trained to understand commands given in natural language. It also analyzes telemetry, provides spoken reports and predicts mission completion.
The neural network was developed with the needs of the agricultural sector in mind. The language model is deployed in a secure cloud environment, making it less vulnerable to external interference and independent of the provider’s decisions.
Particular attention was paid to safety: built-in scenarios and an independent computer vision system ensure that the robot will not harm people or equipment even when it receives incorrect commands.
“When the connection is unstable, the assistant does not freeze, and the system smoothly switches to an onboard local subsystem, which continues to provide spoken updates on the mission and current tasks, while more complex cloud-based forecasts are simply delayed until the connection is restored. In other words, when the connection is lost, only advanced forecasting is unavailable, not control of the robot,” said Gennady Savitsky, lead developer at Cognitive Pilot.
