Abstract:

Using AI in Mining Environment to Improve the Communications in Operating Vehicles (Speech to Text)

Vehicles and driving is one of the highest risks to safety in mining environment. Operating heavy and light vehicles poses a risk of collisions. Communications in between vehicle operators, or the lack thereof, is a factor in many potentially fatal incidents (PFI).

To reduce the risk of miscommunications, which can lead to PFIs or worse, fatalities, vehicle operators must adhere to protocols for positive communications, for example:

Establish radio contact with another vehicle before entering a 50m range, and identify itself by its equipment ID (example: “229 haul truck”)
Request permission to pass the other vehicle, before doing so
Proceed to pass, only after having received permission by the other vehicle

In practice, it is difficult to monitor the adherence to positive communications protocols. Thousands of distinct communications in between vehicle operators occur every day at a mine, resulting in hundreds of hours of recorded audio on the radio communication channels.

In this presentation:

We will demonstrate that AI technology can assist us in monitoring the quality of radio communications between vehicle operators.
We will also show that AI technology (speech to text tool) can be a useful to prevent miscommunication and safely operate vehicles by reducing collision risks.

We will explore the following:

What are the unique challenges that stand in the way of leveraging AI in a mining environment?
Can AI be made to ‘understand’ communications in between vehicle operators (i.e. in between 2 humans)?
Can AI ‘learn’ what is a positive communication by feeding it a large amount of data?
What did we learn after the completion of a proof-of-concept?

Approach

Rio Tinto initiated a proof-of-concept project at its Labrador City site (IOC Iron Ore Company of Canada), encompassing the following:

Extraction of 6 hours of recorded audio of radio communications over night and day shifts
Creation of an AI program with our technology partner
Development of AI components: audio signal processing, speech recognition, natural language processing, and positive communications classification engine
Feeding the AI with a large set of data: thousands of communications and their transcripts
Machine learning: optimization by iteration

Key Findings

In this presentation, we will share the key findings from our proof-of-concept project. We will also suggest how these same findings can be leveraged for future projects whose goals are to use AI technology to safely operate vehicles, by reducing the risk of miscommunication.

Menu

Mr. Glenn Emberley